Middlesex University Research RepositoryAn open access repository of
Middlesex University research
http://eprints.mdx.ac.uk
Sial, Sara Baber (2013) Communicating simulated emotional states of robots by expressivemovements. Masters thesis, Middlesex University. [Thesis]
Final accepted version (with author’s formatting)
This version is available at: http://eprints.mdx.ac.uk/12312/
Copyright:
Middlesex University Research Repository makes the University’s research available electronically.
Copyright and moral rights to this work are retained by the author and/or other copyright ownersunless otherwise stated. The work is supplied on the understanding that any use for commercial gainis strictly forbidden. A copy may be downloaded for personal, non-commercial, research or studywithout prior permission and without charge.
Works, including theses and research projects, may not be reproduced in any format or medium, orextensive quotations taken from them, or their content changed in any way, without first obtainingpermission in writing from the copyright holder(s). They may not be sold or exploited commercially inany format or medium without the prior written permission of the copyright holder(s).
Full bibliographic details must be given when referring to, or quoting from full items including theauthor’s name, the title of the work, publication details where relevant (place, publisher, date), pag-ination, and for theses or dissertations the awarding institution, the degree type awarded, and thedate of the award.
If you believe that any material held in the repository infringes copyright law, please contact theRepository Team at Middlesex University via the following email address:
The item will be removed from the repository while any claim is being investigated.
See also repository copyright: re-use policy: http://eprints.mdx.ac.uk/policies.html#copy
Middlesex University Research Repository:
an open access repository of
Middlesex University research
http://eprints.mdx.ac.uk
Sial, Sara Baber, 2013. Communicating simulated emotional states of
robots by expressive movements. Available from Middlesex University’s
Research Repository.
Copyright:
Middlesex University Research Repository makes the University’s research available electronically.
Copyright and moral rights to this thesis/research project are retained by the author and/or
other copyright owners. The work is supplied on the understanding that any use for
commercial gain is strictly forbidden. A copy may be downloaded for personal, non-
commercial, research or study without prior permission and without charge. Any use of the
thesis/research project for private study or research must be properly acknowledged with
reference to the work’s full bibliographic details.
This thesis/research project may not be reproduced in any format or medium, or extensive
quotations taken from it, or its content changed in any way, without first obtaining permission
in writing from the copyright holder(s).
If you believe that any material held in the repository infringes copyright law, please contact
the Repository Team at Middlesex University via the following email address:
The item will be removed from the repository while any claim is being investigated.
Communicating Simulated Emotional States
of Robots by Expressive Movements
A thesis submitted to Middlesex University in partial fulfilment of the
requirements for the degree of Master of Science (by Research)
Sara Baber Sial (Masters by Research)
School of Science and Technology
Middlesex University
September 2013
©S.B.Sial, 2013
ii
This work is dedicated to
Mr Muhammad Baber Sial- my husband
©S.B.Sial, 2013
iii
Abstract: This research focuses on the non-verbal emotional communication of a non-android
robotic arm used for Human Robot Interaction (HRI). It investigates whether
products, by moving in a life-like way, can communicate their emotions and
intentions to humans or not. The research focuses mainly on the mechanoid robot
(IGUS Robolink) whether it is able to communicate its emotions to the user or not. It
further inspects about the motion parameters that are important to change the
behaviour of mechanoid robot used.
In this study, a relationship is developed between the motion of the robot and the
perceived emotion. The validity of the perceived emotion by the user is later
checked using three different emotional models: Russell’s circumplex model of
affect, Tellegen-Watson-Clark model and PAD scale. The motion characteristics
such as velocity and acceleration are changed systematically to observe the change
in the perception of affect caused by the robotic motion. The perceived affect is then
marked by the user on all three emotional behaviour models.
The novelty of the research lies in two facts: Firstly the robotic embodiment used
does not have any anthropomorphic or zoomorphic features. Secondly the
embodiment is programmed to adopt the smooth human motion profile unlike
traditional trapezoidal motion used in industrial robots.
From the results produced it can be concluded that the selected motion parameters of
velocity and acceleration are linked with the changed of perceived emotions. The
emotions at low values of motion parameters are perceived as sad and unhappy. As
the values for motion parameters are increased the perceived emotion changes from
sad to happy and then to excited. Moreover the validity of perceived emotions is
proved as the emotion marked by the user is same on all the three scales, also
confirming the reliability of all the three emotional scale models. Another major
finding of this research is that mechanoid robots are also able to communicate their
emotions to the user successfully. These findings for Human-Robot interaction on
user’s perception of emotions are important if robots are to co-exist with humans in
various environments, such as co-workers in industry or care-workers in domestic
settings.
©S.B.Sial, 2013
iv
Acknowledgments: I would like to thank Prof. Mehmet Karamanoglu, Head of Department of Design
Engineering and Mathematics, who helped me in getting through all the process of
finalizing my admission and introducing me to this project as well as Dr Aleksandar
Zivanovic, my supervisor, who has helped me in achieving all the objectives that
were needed to complete this project.
Thanks to all other who are not mentioned here but have supported me in this
project and my stay over here. They will be always in my memories.
Last but not the least I would like to thank my husband, Muhammad Baber Sial, my
Parents and Dr.Irtiza Ali Shah (NUST), for constant support and making me believe
that everything can be achieved if you just put your sincere efforts into it.
©S.B.Sial, 2013
v
Table of contents
ABSTRACT…………………………………………………………………………....iii
ACKNOWLEDGEMENTS ……………………………………………………..........iv
TABLE OF CONTENTS………………………………………………………… …...v
LIST OF FIGURES………………………………………………………………….viii
LIST OF TABLES…………………………………………………………………….xi
CHAPTER 1- INTRODUCTION………………………………………................................................1
1.1 DEFINITION OF HRI…………………………………………………………………….....1
1.2 TYPES OF ROBOTS………………………………………………………………………..3
1.3 SOCIALLY EVOCATIVE ROBOTS……………………………………………………….4
1.4 NATURAL MOVEMENTS OF HUMANS VERSUS INDUSTRIAL ROBOTS………….4
1.5 COMERCIAL APPLICATION IN INDUSTRY…………………………………………....5
1.5.1 MINERVA museum tour-guide robot……………………………………….....6
1.5.2 Nursing robots………………………………………………………………….6
1.5.3 NASA humanoid robots………………………………………………………..7
1.5.4 KISMET………………………………………………………………………..7
1.6 COMPARISON OF DIFFERENT ROBOTS…………………………………………….....8
1.7 SUMMARY OF THE CHAPTER…………………………………………………………..9
CHAPTER 2- AFFECTIVE EXPRESSIONS OF MACHINES…………………...10
2.1 EXPRESSION OF EMOTIONS………………………………………………...................10
2.2 TYPES OF COMMUNICATION………………………………………………………….10
2.3 HOW MACHINES EXPRESS EMOTIONS………………………………………………11
2.4 USER PERCEPTION OF EMOTIONS……………………………………………………12
2.5 SUMMARY OF THE CHAPTER………………………………………………………….12
CHAPTER 3-HARDWARE AND SOFTWARE PLATFORM FOR
RESEARCH………………………………………………………………………………………………………………………..13
3.1 INTRODUCTION OF HARDWARE “IGUS-ROBOLINK”…............................................13
3.2 FEATURES OF IGUS-ROBOT USED FOR HRI………………………………………....14
3.3 SPECIFICATION OF ROBOTIC ARM USED…………………………………………....15
3.4 KINEMATICS OF ROBOT…...............................................................................................15
3.5 INTRODUCTION TO LABVIEW…………………………………………………………16
3.5.1 Reasons for using LabVIEW…………………………………………………….17
3.6 SELECTION OF NI HARDWARE PLATFORM…………………………………………18
3.6.1 CompactRIO……………….…………………………………………………….18
3.6.2 Stepper drivers 9501……….…………………………………………………….19
3.7 CONNECTING cRIO 9074 AND IGUS-ROBOLINK ……………………………………20
3.7.1 Hardware required……………………………………………………………….20
3.7.2 Software required………………………………………………………………...21
3.7.3 Hardware connections…………………………………………………………21
3.8 SUMMARY OF THE CHAPTER…………………… ……………………………………20
©S.B.Sial, 2013
vi
CHAPTER 4- DESGIN OF ALGORITHM USING LABVIEW………………….23
4.1 INTRODUCTION TO DESIGN OF ALGORITHM………………………………………23
4.2 NI LABVIEW FPGA MODULE…………………………………………………………...23
4.3 NI LABVIEW REAL TIME MODULE……………………………………………………25
4.3.1 Basic real-time architecture……………………………………………………....26
4.3.2 Basic real-time toolkits…………………………………………………………...26
4.3.3 Steps for development of real-time system………………………………............26
4.4 NI LABVIEW SOFTMOTION MODULE………………………………………………...27
4.5 NI LABVIEW ROBOTIC MODULE……………………………………………………...29
4.5.1 Stages involved in implementing inverse kinematics…………………………....31
4.6 SUMMARY OF THE CHAPTER…………………… ……………………………………32
CHAPTER 5- CORE RESEARCH METHODOLOGY FOR PERCEPTION OF
ROBOT EMOTIONS………………………………………………………………....33
5.1 CORE CONCEPT OF RESEARCH………………………………………………………..33
5.2 MODELLING MACHINE EMOTIONS…………………………………………………...34
5.3 EMOTIONAL MODELS FOR RESEARCH………………………………………………35
5.3.1 Russell’s circumplex model of affect…………………………………………….35
5.3.3.1Examples of effective interaction………………………………………38
5.3.2 Tellegen-Watson-Clark model…………………………………………………...38
5.3.3 PAD scale………………………………………………………………………...40
5.4 SELECTION OF GESTURES……………………………………….……………………..42
5.4.1 Graphical illustration of gestures….……………………………………………..42
5.5 MOTION CHARACTERISTICS…………………………………….…………………….44
5.6 SUMMARY OF THE CHAPTER………………………………………………………….47
CHAPTER 6- EXPERIMENTS AND RESULTS…………………………………..48 6.1 EXPERIMENTS FOR EMOTIONAL COMMUNICATION……………………………...48
6.2.1 Experiment procedure……………………………………………………………48
6.2 QUESTIONNAIRE FOR MEASURING PERCEPTION OF EMOTIONS……………….49
6.2.1 Questionnaire for Russell’s model……………………………………………….50
6.2.2 Questionnaire for Tellegen-Watson-Clark model………………………………..51
6.2.3 Questionnaire for PAD model……………………………………………………52
6.2.4 Measurement of emotions by participants………………………………………..53
6.3 EMOTION RECOGNITION BASED ON SCALES………………………………….........56
6.4 MODEL RESULTS………………………………………………………………………....57
6.4.1 Results for Russell’s model………………………………………………………57
6.4.2 Results for Tellegen-Watson-Clark model……………………………………….60
6.4.3 Results for PAD model…………………………………………………………...63
6.5 DISCUSSION OF RESULTS………………………………………………………………67
6.6 SUMMARY OF THE CHAPTER………………………………………………………….72
CHAPTER 7- CONCLUSION AND RECOMMENDATIONS……………………………….73
7.1 CONCLUSION……………………………………………………………………………..73
7.2 RECOMMENDATION REGARDING HARDWARE…………………………………….74
7.3 RECOMMENDATION REGARDING LABVIEW PROGRAMMING…………………..74
7.4 RESEARCH LIMITATIONS………………………………………………………………74
7.5 FUTURE WORK…………………………………………………………………………...75
©S.B.Sial, 2013
vii
REFERENCES………………………………………………………………………..76
APPENDICES
Appendix A: Drawings for IGUS robotic arm…………………………………………86
Appendix B: Tube length of joints……………………………………………………..89
Appendix C: Technical data for tube lengths…………………………………………..90
Appendix D: Datasheet of integrated Hall IC’s and configuration of sensor lines…….92
Appendix E: Technical data for stepper motors………………………………….….....96
Appendix F: Complete specifications for drive unit….................................................103
Appendix G: Technical Datasheet of NI 9501………………………………………..107
Appendix H: Technical Datasheet of cRIO 9074…………………………………….110
Appendix I: Technical Datasheet of NI 9401………………………………………...118
Appendix J: Properties of CompactRIO 9074………………………………………..123
Appendix K: Connection of motors with 9501……………………………………….125
Appendix L: Wiring of chassis and cRIO…………………………………………….126
Appendix M: Mechanical parts and joint types of Robolink…………………………134
Appendix N: Ethical approval and consent form……………………………………..146
Appendix O: LabVIEW Code………………………………………………......CDROM
©S.B.Sial, 2013
viii
List of figures Figure1. 1 : PKD, Tron-X, TOPIO android robots ....................................................................... 3
Figure1. 2: Famous zoomorphic robots ........................................................................................ 4
Figure1. 3: Typical motion of industrial robot .............................................................................. 5
Figure1. 4: Natural human movements ......................................................................................... 5
Figure1. 5: MINERVA famous tour guide robot .......................................................................... 6
Figure1. 6: RIBA- a famous nursing robot ................................................................................... 7
Figure1. 7: NASA humanoid robot ............................................................................................... 7
Figure1. 8: KISMET used for HRI research ................................................................................ 8
Figure2. 1: IGUS robotic arm ..................................................................................................... 11
Figure3. 1: IGUS articulated arm with labelling of different parts ............................................. 14
Figure3. 2: Research platform used ............................................................................................ 15
Figure3. 3: Usage popularity of LabVIEW compared with other software ............................... 17
Figure3. 4: CompactRIO platform .............................................................................................. 18
Figure3. 5: CompactRIO 9074 .................................................................................................... 19
Figure3. 6: Stepper driver 9501 .................................................................................................. 19
Figure3. 7: Complete architecture with NI 9501 ........................................................................ 20
Figure3. 8: NI module 9401 ........................................................................................................ 21
Figure4. 1: FPGA programming palette ..................................................................................... 24
Figure4. 2: FPGA based hardware offered by NI ....................................................................... 24
Figure4. 3: Process of code deployment on FPGA ..................................................................... 25
Figure4. 4: Basic architecture ..................................................................................................... 26
Figure4. 5: RT toolkit ................................................................................................................. 26
Figure4. 6: Summary of development process ............................................................................ 27
Figure4. 7: SoftMotion palette for programming ........................................................................ 27
Figure4. 8: Spline generation function ........................................................................................ 28
Figure4. 9: Spline generation loop .............................................................................................. 28
Figure4. 10: Detailed process for Spline generation ................................................................... 29
Figure4. 11: Robotics toolkit ...................................................................................................... 29
Figure4. 12: Function of inverse kinematics ............................................................................... 30
Figure4. 13: Parameters for inverse kinematics of robot ............................................................ 30
Figure4. 14: Serial arm with 5 revolute joints ............................................................................ 31
Figure5. 1: Emotional vehicles responding to light .................................................................... 34
©S.B.Sial, 2013
ix
Figure5. 2: Computational model for emotional vehicle ............................................................ 35
Figure5. 3: Russell’s circumplex model of emotions ................................................................. 36
Figure5. 4: Russell’s circumplex model for affective interaction ............................................... 38
Figure5. 5: Tellegen-Watson-Clark model ................................................................................. 39
Figure5. 6: PAD model ............................................................................................................... 42
Figure5. 7: Graphical representation of point-point motion ....................................................... 43
Figure5. 8: Graphical representation of waving of robot ............................................................ 43
Figure5. 9: Graphical representation of bowing of robot ............................................................ 44
Figure5. 10: Spline curve for G1 at V=250 ................................................................................ 45
Figure5. 11: Spline curve for G1 at V=800 ................................................................................ 45
Figure5. 12: Spline curve for G1 at V=2000 .............................................................................. 45
Figure5. 13: Spline curve for G2 at A=15 .................................................................................. 46
Figure5. 14: Spline curve for G2 at A=5 .................................................................................... 46
Figure5. 15: Spline curve for G2 at A=1.5 ................................................................................. 46
Figure6. 1: Russell's model questionnaire................................................................................... 50
Figure6. 2: Tellegen-Watson-Clark model questionnaire ........................................................... 51
Figure6. 3: PAD questionnaire.................................................................................................... 52
Figure6. 4: Russell's questionnaire filled by the participant ....................................................... 53
Figure6. 5: Tellegen-Watson-Clark questionnaire filled by the participant ................................ 54
Figure6. 6: PAD questionnaire filled by the participant ............................................................. 55
Figure6. 7: Russell’s model graph for 3 sets of parameters for G1 ............................................ 57
Figure6. 8: Russell’s model graph for 3 sets of parameters for G2 ............................................ 58
Figure6. 9: Russell’s model graph for 3 sets of parameters for G3 ............................................ 59
Figure6. 10: Tellegen-Watson-Clark model graph for 3 sets of parameters for G1 ................... 60
Figure6. 11: Tellegen-Watson-Clark model graph for 3 sets of parameters for G2 ................... 61
Figure6. 12: Tellegen-Watson-Clark model graph for 3 sets of parameters for G3 ................... 62
Figure6. 13: PAD model graph for 3 sets of parameters for G1 ................................................. 63
Figure6. 14: PAD model graph for 3 sets of parameters for G2 ................................................. 65
Figure6. 15: PAD model graph for 3 sets of parameters for G3 ................................................. 66
©S.B.Sial, 2013
x
List of tables
Table1. 1: Comparison of different robots .................................................................................... 8
Table3. 1: DH parameters of IGUS robot 5DOF ........................................................................ 16
Table5. 1: Location of emotion on circular graph....................................................................... 37
Table6. 1: Response of participants for Russell’s model G1 ...................................................... 57
Table6. 2: Response of participants for Russell’s model G2 ...................................................... 58
Table6. 3: Response of participants for Russell’s model G3 ...................................................... 59
Table6. 4: Response of participants for Tellegen-Watson-Clark model G1 ............................... 60
Table6. 5: Response of participants for Tellegen-Watson-Clark model G2 ............................... 61
Table6. 6: Response of participants for Tellegen-Watson-Clark model G3 ............................... 62
Table6. 7: Response of participants for PAD model G1 ............................................................. 63
Table6. 8: Response of participants for PAD model G2 ............................................................. 64
Table6. 9: Response of participants for PAD model G3 ............................................................. 66
1
CHAPTER 1
Introduction
Currently, most robots in industry work independently of humans, due to safety
concerns. Robots typically operate in work cells with safety fences surrounding
them which shut down the robot if a person enters.
Researchers are now trying to develop effective interaction of robots with humans
for different purposes such as entertainment, medical diagnosis, exchange of
information and much more. As the robots can move anywhere around offices,
hospitals and homes, they need to interact safely with humans rather than being an
obstruction or a danger.
In a situation where a robot can choose one of multiple options to achieve its goal,
the next movement of that robot might not be clear to a human. For this reason it
may cause an accident or it might not be safe to work with robots if a human is not
aware of its intentions for the next move. For this reason human-robot interaction
is important as one can infer the intention of the robot if the motion is interactive.
The central focus of this thesis is to develop safe human-robot interaction in social
environments. This thesis also discusses various emotional models that are
relevant for HRI. A design algorithm is proposed for interaction of IGUS robot
and guidelines to improve this algorithm for interaction. Several results are
presented based upon the experimentation keeping in view the various factors that
affect HRI.
1.1 Definition of HRI
M. A. Goodrich and A. C. Schultz (2007) define Human-robot interaction as “A
field of study dedicated to understanding, designing, and evaluating robotic systems
for use by or with humans”.
This human-machine interaction is usually non-verbal communication. As by
definition “Non-verbal communication serves as a rich source of information in inter
human communication” (Saerbeck and Bartneck, 2010). As the motion in itself
contains a lot of information, one can easily predict the physical state intention from
the robot’s motion. One can relate this non-verbal human-robot interaction with
©S.B.Sial, 2013
2
human-animal interaction. Although animals cannot speak human language, or
cannot interact with them verbally, but from their gestures and motion they can tell
humans their different states of emotions that include happiness, anger, sadness,
boredom, hunger and many more.
The “Success of a robotic platform depends upon more than mere task
performance.” (Saerbeck and Bartneck, 2010). For example, if the robot is
programmed for speedy cleaning with fast performance, humans might perceive it as
angry or aggressive. So in order for successful and complete interaction with robots,
it is necessary to understand how humans perceive their motion and behaviour. The
research described in this dissertation focuses on designing an algorithm in
LabVIEW that helps the robot to develop and produce expressive and interactive
movements to communicate with humans.
Robots now in the market are introduced as co-workers such as KUKA Roboter
GmbH (Haddadin et al., 2011) and Baxter (Anandan, 2013) etc. The rapid growing
market of HRI give rise to different types of robots. Some developers and
researchers believe that humanoid robots are important for natural and effective
interaction. As defined by (Bartneck et al., 2006) that “Designing androids with
anthropomorphized appearance for more natural communication encourages a
fantasy that interaction with robots is thoroughly human like and promotes
emotional or sentimental attachments”.
Anthropomorphism is a term that is widely used in the robotics world.
“Anthropomorphism refers to the attribution of a human form, human
characteristics, or human behaviour to non-human things such as robots, computers
and animals”, (Bartneck et al., 2009). Research has shown that if the interface is
humanoid the expectations of humans increases tremendously such that the robot
might not be able to fulfil them, while for the machine interface, the level of
expectations from robots is lowered (Bartneck et al., 2006). The possible
explanation for this situation could be that people look with different aspects
towards human and robots. According to the Mori’s Uncanny valley theory (Mori,
2005) the degree of empathy increases as the robot becomes more human-like.
©S.B.Sial, 2013
3
1.2 Types of robots
Android
Machine robots
These are two main kinds of robots used in the industry and for research. The robots
that look like a machine without any anthropomorphic or zoomorphic features are
called as machine robots. Robots, especially androids, are being developed as a
helpers and co-workers as well for commercial purpose where they are used as toys
and for other household purposes. These robots look like humans or animals. They
are further divided in the categories of anthropomorphic and zoomorphic robots that
are explained below:
Anthropomorphic robots have human like appearance like facial expressions,
humanoid head mounted on a neck with eyes and ears, skin etc. These are also
known as humanoid robots. Some popular anthropomorphic robots are: PKD, Tron-
X, TOPIO ("TOSY Ping Pong Playing Robot") and many more. These robots are
shown in Fig. 1.1
Another type of android robot are one that looks like animals (Boris, 2003).
Zoomorphism refers to the shape of something in form of animals. These robots
raise the level of expectations because of their appearance as compared with
machine interface robots. Some of the famous zoomorphic robots are: Sony AIBO,
Lamprey etc. These can be seen below in Fig. 1.2.
Research and questionnaires have shown that people look differently towards
anthropomorphic and zoomorphic robots. Bartneck et al., (2006) have carried out
Figure1. 1 : PKD, Tron-X, TOPIO android robots (dick, 2013)
Fig. 1.1(a) shows a PKD robots sitting and staring, Fig. 1.1(b) shows Tron-X interacting with
environment and Fig. 1.1(c) shows TOPIO playing table tennis
(a) (b) (c)
©S.B.Sial, 2013
4
research in which they have concluded that “Because ABIO is a zoomorphic robot,
not a humanoid, we believe that people did not expect it to demonstrate a very good
performance on task”.
1.3 Socially evocative robots
Traditionally the term ‘social robots’ was used for multiple robots working together.
There are a lot of challenges that are faced in the research of Human-Robot
interaction in terms of nature of interaction and social behaviour (Dautenhahn,
2007). In today’s research world this term is usually used to differentiate between
human-interactive anthropomorphic robots from other type of robots (Breazeal,
2003). Recent commercial and industrial applications are emerging where human
robot interaction is an important aspect of robotics.
There are several subclasses of social robots such as anthropomorphic, zoomorphic,
caricatured and functional (Fong et al, 2003). Entertainment robots like AIBO,
Furby etc. are well-known. Similarly Lego Mind-storm kits are popular but are
aimed more at the educational market.
The interactive capabilities of these robots are limited, but this is quite motivational
for carrying out further research in this area. These are socially evocative, socially
communicative, socially responsive, and sociable as described by Breazeal in her
paper (Breazeal, 2003).
1.4 Natural movements in human versus industrial robots
Most industrial robots focus only at the high precision of the end effector reaching
the target position and typically, use a trapezoidal velocity profile (see Fig. 1.3).
Figure1. 2: Famous zoomorphic robots (Gizmag, 2002)
Fig.1.2 (a) represents AIBO that resembles a dog and Fig.1.2 (b) represents a zoomorphic
robot that resembles a snake
(a) (b)
©S.B.Sial, 2013
5
Fig. 1.3 shows that the robot motion profile is quite sharp. So if the robots have to
work as co-workers with humans in industry, the motion is quite unpredictable,
because it appears jerky and “un-natural” (Flanagan et al., 1990). It is proposed that
if a robotic system adopted a human profile for its motion, it would appear to move
more “naturally”, which might make it safer to work together with humans
(Gaertner et al., 2010). The velocity motion profile of human limb movement is a
bell shaped-smooth curve without the sharp edges seen in the profile of industrial
robots. Natural human movements for position, velocity and acceleration profile are
shown in Fig. 1.4(Gaveau and Papaxanthis, 2011).
The process of trajectory formation in human arm movements is more complex than
simply alerting between the equilibrium positions. For example it is proved that if
the arm is displaced from its normal trajectory during movements, it will not return
to initial or final equilibrium positions but will move to points intermediate between
them (Bizzi et al., 1984).
Figure1. 3: Typical motion of industrial robot (Sandin, 2003)
Figure1. 4: Natural human movements (Gaveau and Papaxanthis, 2011)
©S.B.Sial, 2013
6
The aim of this research is to program a low-level control system for the given
robotic platform in order to develop a motion profile that resembles human’s motion
as in Fig. 1.4.
1.5 Commercial applications of robots
Commercial robotic applications are now widely using robot-robot and human-robot
interaction technologies (Koeppe et al., 2003). Although the capability of android
robots is restricted in terms of interaction with humans, it is a rapidly growing field
of research. These robots are not only used for entertainment purposes but also have
several other applications in industry. Some of the popular interactive robots are as
shown in Fig. 1.5 to 1.8.
1.5.1 MINERVA museum tour-guide robot
This is a popular interactive tour-guide robot used in the Smithsonian museum.
During the interaction of two weeks, it met thousands of different people traversing
more than 44 km at speeds of up to 163 cm/sec (Thrun et al., 1999). The purpose of
robot was to describe the exhibits to visitors.
1.5.2 Nursing robots
Robots are now increasingly used for nursing and caring applications. The tasks of
these robots include helping the elderly to move around in a room, taking them to
toilets, helping them to lay down etc. RIBA is one of the well-known examples of
nursing robots which resembles a friendly bear.
Figure1. 5: MINERVA famous tour guide robot (Thrun et al., 1999)
©S.B.Sial, 2013
7
1.5.3 NASA humanoid robot
NASA developed a humanoid robot which acts as an assistant for astronauts. The
fame of these robots is not only because of their ability for carrying out their task but
also in the fact of how they interact and behave with people around them. Fig. 1.7
shows NASA humanoid robot.
1.5.4 KISMET
Kismet (see Fig. 1.8) is a robot that is used for human-robot interaction. It has got
various input features for interacting with human beings. It can produce several
facial expressions, voices and other actions. To produce facial expressions it has got
eyebrows, lips, jaws and various other features.
Figure1. 6: RIBA- a famous nursing robot (Uno, 2009)
Figure1. 7: NASA humanoid robot (NASA, 2013)
©S.B.Sial, 2013
8
Figure1. 8: KISMET used for HRI research (Menzel, 2013)
1.6 Comparison of different robots
There are different kinds of robots available for various purposes. The usefulness of
the robot depends on how they interact and sense the environment and surroundings
around them (Jason, 2007).The design of robot influences the human-robot
interaction significantly (Forlizzi and DiSalvo, 2006). Table 1.1 shows different
kinds of robots that are used in industry, entertainment and in several other fields.
Robots Type of robot Use
KISMET Android Used for HRI, produce various facial expressions
NASA Humanoid Acts as an assistant for astronauts
Geminoid TMF Humanoid Mimics a person’s facial expressions
MOTOMAN Robotic arm Industrial robot used for painting
AIBO Zoomorphic Used for interaction and entertainment
SCARA Robotic arm Used for industrial purposes
Micro Flying Robot Mechanoid Used as a flying camera
MINERVA Mobile Used as a tour guide in museum
PUMA Robotic arm Used for industrial purposes
RIBA Zoomorphic Used for nursing purposes
Robocup Humanoid Used for playing football
Table1. 1: Comparison of different robots
©S.B.Sial, 2013
9
1.7 Summary of the chapter
This chapter reviews the robots that are socially interactive and highlights the
importance of human-robot interaction in today’s world. There are several types of
robots discussed in this chapter. It discusses how a robot can have “natural”
movements that can be anticipated by the humans. It then compares various different
types of robots and there use in industry as well as domestic fields.
The next chapter discusses how these robots can be used for the effective
communication by their expressive movements.
©S.B.Sial, 2013
10
CHAPTER 2
AFFECTIVE EXPRESSIONS OF MACHINES
2.1 Expression of emotion
Recent research in human-robot interaction has shown that emotions play an
important role in designing any interface, as the machines are now perceived as
social actors (Nass, 1996). As explained by (Picard, 1997b), people are usually seen
expressing their frustration to computers when they are not working by shouting or
yelling at them.
However, the distinctness of the expression depends strongly on the type of
embodiment. So if the embodiment is a humanoid rather than machine interface
robot, it will express its emotions more prominently. Research in the field of HRI
shows that these emotional capabilities play a significant role in decision making
(Barnes, 1996) and problem solving (Fesit, 1994).
2.2 Types of communication
There are mainly two kinds of communication:
Verbal communication
Non-verbal communication
Communication that involves speech is called verbal communication and is a natural
way for humans to express their emotions. Thrun et al., (1999) states that “The most
influential parameters for emotional expression in speech is pitch (level, range and
variability)”. Humanoid robots are usually capable of verbal communication, but
what if the robots cannot express their emotion by speech? Body language and
gestures are considered to be an important aspect for expression of emotion (Thrun
et al., 1999).
The main emphasis of this thesis is on non-verbal and behavioural communication,
as the type of embodiment used for this project is a robotic arm without the
capability of verbal communication. Mime artists are good example of non-verbal
communication through their body gestures and facial expressions. Body gestures
©S.B.Sial, 2013
11
are an important aspect as the embodiment used in this research does not have any
anthropomorphic or zoomorphic features for expressing its emotions.
2.3 How machines express emotions
The main focus of this research is on designing expressive behaviours of robots that
have machine interface like IGUS robotic arm shown in Fig. 2.1. The reason for
choosing this specific platform of mechanoid robot rather than some industrial
embodiment is because one can get into the low level programming of this kind of
robot and make it move in a way that is natural and closely resembles with human
movements. The robots are now widely used to express their emotions by movement
(Matsumaru, 2009). Physical movements hold great importance for the emotional
interaction between humans and products (Qassem at el., 2010).
The feature of being interactive for this robotic arm actually means that it should
exhibit some expressive movements based on various parameters, as expressive
movements are the main content for non-verbal communication. People usually
interpret motion pattern based on emotions (Heider and Simmel, 1944). The motion
pattern and movement trajectory of the robot plays a substantial role in how a user
perceives its emotions. This movement of the robot is actually interpreted as an
emotional behaviour. There are several motion features that are the cause of
Figure2. 1: IGUS robotic arm (Fontys, 2013)
©S.B.Sial, 2013
12
expressing emotions (Saerbeck and Van Breemen, 2007). According to the
hypothesis if we change these motion parameters, the perceived emotions for the
particular embodiment is also changed.
2.4 User Perception of emotions
User perceptions of the emotions of robots depend on several factors. Research in
the field of HRI has introduced several models that are used for perception of
emotions. An evolutionary model suggests that the ability to correctly judge the
emotions and to perceive intentions correctly is important in order to integrate robots
effectively in everyday life (Saerbeck and Bartneck, 2010). For example one can
easily tell the emotions of a mountain lion by observing it and say whether it is
angry, hungry for prey, relaxed, mating or wandering (Blythe et al., 1999).
According to another model, social reasoning also contributes a lot towards the
emotional behaviour (Wondolowski and Davis, 1991).
However, anthropomorphism also contributes towards the perception of emotions. If
the robot has high anthropomorphism (i.e. very close resemblance to humans), the
expectations of the user are high for this robot as compared to zoomorphic. However
comparing a zoomorphic and machine like robot, the expectations are further
lowered with a robot that looks like a machine. This behaviour of user perception for
emotions is explained by the Uncanny Valley theory (Mori, 2005). The factors used
in this research to animate various emotions are mainly velocity, acceleration and
spline motion of the joints of the robot. These features and the scales used for them
will be discussed in later chapters.
2.5 Summary of the chapter
This chapter discusses how a robot can use it’s movements to express the emotions
and describes the reason for choosing a mechanoid platform rather than an industrial
robot. Effect of motion parameters on the perceived behaviour is discussed in this
chapter. It highlights the fact that how a user will perceive the emotions in machines
according to Uncanny Valley theory and how the level of expectation is linked with
the type of embodiment.
The next chapter will focus in detail on the hardware and software platform that is
used for expressing the emotional behaviour of machines for this research.
©S.B.Sial, 2013
13
CHAPTER 3
HARDWARE AND SOFTWARE PLATFORM
FOR RESEARCH
3.1 Introduction of hardware “IGUS-Robolink”
IGUS is a German company specializing in plastic bearings and cable management
systems etc. One of their products is the Robolink system, which is a range of
configurable joints and links that allows customers to specify the number of joints,
lengths of links, etc. The joints are actuated by flexible cables which are routed
through the hollow links (IGUS, 2012a). They produce four different types of
joints. Optionally, incremental encoders may be specified that are used for tracking
the position of joint. The system is not supplied with end effectors, but different
types of end effectors can be fitted at the end of the last plastic link, like cameras,
grippers, light actuators etc. The articulated arm is designed in a way that the cables
for these actuators can be routed through the body of robot.
The Robolink system is basically a toolbox of mechanical components that can be
put together to make a robotic arm. This product was launched three years ago and
the first mechanical component of this toolbox was a plastic link with tendon drive
(IGUS, 2009).
The main parts of this robotic arm are: stepper drives, drive units for these motors, a
cable system to deliver motion in the articulated arm, incremental encoders, plastic
joints and rigid links. The features and working of each of these will be discussed in
detail.
The particular articulated arm that was used for this project had 5DOFs, with three
rotational and two pivot joints. It had three link rods and also had incremental
encoders. Fig. 3.1 represents the different parts of the IGUS articulated robotic arm.
©S.B.Sial, 2013
14
A. Joint
B. Robolink bowden cable
C. Robolink multi axis joint
D. Robolink connecting tubes
E. Robolink flange shaft support
F. Movement through pulleys
G. IGUS stepper motor
H. Robolink drive wheel
I. Housing drive unit
J. Dyneema ropes
3.2 Features of IGUS-Robolink used for HRI
There are several features of the Robolink system that makes them suitable for the
reasearch described in this dissertation.
The joints in these articulated arms are made of polymers. The reason for using
plastic is that it is light in weight (only 350 grams), the joints do not need any kind
of lubrication, they are low in price and have longer life.
These arms are compact because each joint unit has 2 DOF, one pivoting and one
rotational. Also the link length is configurable because the links are simple tubes.
Figure3. 1: IGUS articulated arm with labelling of different parts (IGUS, 2013)
©S.B.Sial, 2013
15
The tendon drive system for these robotic arms makes it easy for the designer to
choose the drive and control elements freely. For position accuracy and precision,
angular positioning indicators can also be ordered with these arms that have a
precision of 0.07 degrees.
Another feature of this robotic arm is that the drive system is freely selectable. The
joints are driven by a flexible sinews (rope) system. The tightness of these ropes
drive is adjustable. Moreover alternate drive or control systems are very easy to
introduce in these articulated arms. Stepper motors were used in the system used for
this research. The detail on mechanical parts and joint types of the IGUS-Robolink
are attached in appendix M.
3.3 Specification of robotic arm used
The weight of the arm is 350g including the plastic joints, connecting tubes and
ropes. The joint is made up of fine polyamide 2200 (IGUS, 2013.). The specific arm
used in this research has a part number of RL-50-DOF5-28-WS. The component
number for this robotic arm is TL-002-001. 001 and 002 represents the joint
versions. Rl-50-002 WS is the one with angle sensors and rotation allowed by this is
+130/-50°. Whereas for Rl-50-001 the rotation allowed is +/-90°. WS in the product
code indicates that the joints are equipped with angle sensors. DOF5 represents that
it has 5 degrees of freedom, with the base joint as a rotational. Of the remaining
joints, two are pivot and two are rotational joints. There are three 0.4m links. The
specifications of particular arm that is used for research is stated below.
The arm that is used in the research of this project has 5 DOFs. Fig. 3.2 represents
the DOF for specific arm that is being used:
Figure3. 2: Research platform used (Fontys, 2013)
©S.B.Sial, 2013
16
3.4 Kinematics of robot
Denavit-Hartenberg method was used for calculating the kinematics of the IGUS
robot that is used in the research. This method is widely used to determine the direct
kinematics of robot by specifying some of the parameters. According to DH
conventions the coordinate frame of link i+1 with respect to the coordinate frame of
link i can be represented by following matrix (NI, 2013).
DH parameters calculated for IGUS robot are shown in Table 3.1 below:
Link Number Length
(meters)
Twist angle
(radians)
Offset distance
(meters)
1 0.4 0 0
2 0 -1.5 0
3 0.4 0 0
4 0 1.5 0
5 0.4 0 0 Table3. 1: DH parameters of IGUS robot 5DOF
3.5 Introduction to LabVIEW
LabVIEW stands for Laboratory Instrument Engineering Workbench. This system
was developed when National instruments started to look for some way by which
they could reduce the time that is required to program instrumentation systems
(Travis and Kring, 2013).This graphical programming language is used in academic,
research, industry and many more fields.
It is multi-purpose software that can be used for testing and measurement,
monitoring, simulation and for process control and automation. Its popularity is due
to unparalleled connectivity to instruments, powerful data acquisition capabilities,
natural dataflow based graphical programming interface, scalability, and overall
function completeness (NI, 2011a).
LabVIEW has the capability of running on multiple devices. The coding is done by
the user in an environment provided by LabVIEW software and then it is deployed
on the target. Some of the commonly used targets in LabVIEW are CompactRIO
which are basically programmable automation controllers, programmable device
©S.B.Sial, 2013
17
arrays (PDA’s), real time operating system (PXI), microcontrollers, or field
programmable gate arrays (FPGAs)(Folea,2011).
There are many built-in libraries, examples, software drivers for data acquisition that
are available in LabVIEW. LabVIEW has toolkits used for signal processing, data
analysis, mathematics, real-time programming, simulation, robotics and many more.
The popularity and expansion of LabVIEW in market, academics, research and other
fields for engineering, design, simulation, and testing etc. can be seen in Fig. 3.3.
According to the research by the producers of LabVIEW (NI, 2009) “In 2004,
National Instruments measurement hardware provided customers with more than
6,000,000 virtual instrumentation measurement channels. From low cost USB data
acquisition, to process control vision systems and image acquisition, to RF
measurement at 2.7GHz, to GIPB bus communication, National Instruments has
shown more than 25,000 companies that it offers the measurement hardware and
scalable hardware platform required to complete virtual instruments”
3.5.1 Reasons for using LabVIEW
Some of the reasons why this software is chosen over others for research purposes
are: (Ertugrul, 1999):
It allows the user to develop his/her own virtual environment for
programming and provides a user-friendly interface that is economical and
adaptable
Figure3. 3: Usage popularity of LabVIEW compared with other software (NI, 2013b)
©S.B.Sial, 2013
18
It has some multimedia ability which makes it even more friendly (e.g.
adding voices, warnings etc.)
It generates a report file at the end in notepad format that is easy to
understand by the user
Its capable of printing a specific part of the program that user wants
Capability of placing access limitations for others on different parts of the
code
Can link to other popular software like Pro-E etc.
Provides many examples, tutorials and test programs
Beside all these, it provides the user with the option of thousands of modules and
toolkits of robotics, instrument and measurement, kinematics etc.
3.6 Selection of NI hardware platform
There are many hardware platforms available from National Instruments that can be
used for the purpose of research for connecting with IGUS-Robolink. The one
selected is CompactRIO 9074, which is a high performance programmable
automation controller (PAC).
3.6.1 CompactRIO
“CompactRIO is a reconfigurable embedded control and acquisition system” (NI,
2013a). The hardware platform of CompactRIO contains slots for various
input/output modules, a reconfigurable FPGA chassis, with an embedded controller.
It can be used with LabVIEW for a variety of different applications like
measurement and testing, robotics, embedded control etc.
Figure3. 4: CompactRIO platform (NI, 2013a)
©S.B.Sial, 2013
19
The specific CompactRIO used for this research is cRIO 9074 see Fig. 3.5, with the
following features (NI, 2012a) as stated below. The datasheet is attached in the
appendix H:
This is an integrated system that combines a real time processor and a
reconfigurable Field Programmable Gate Array (FPGA) within the same chassis.
The real time processor is 400MHz with 2M gate FPGA. There are eight slots
available in the chassis for different input output modules. The DRAM provided by
the system is 128MB for embedded operations and 256MB of non-volatile memory
for data logging. For network programming, communication it is provided by two
10/100 Mb/s Ethernet ports. Properties of the CompactRIO used in this research are
in appendix J.
3.6.2 Stepper drivers 9501
NI 9501 is a C-series stepper driver that can be used with cRIO 9074 to operate
stepper motors used in the IGUS Robolink. The datasheet for this module is in
appendix G.
Figure3. 5: CompactRIO 9074 (NI, 2012c)
Figure3. 6: Stepper driver 9501(NI, 2012c)
©S.B.Sial, 2013
20
This driver is equipped with all of the features to control and power the stepper
motor. It is capable of interfacing with the FPGA in the chassis, and then controlling
the stepper motors by step and direction in programming. Fig. 3.7 shows the
complete architecture of NI 9501 with CompactRIO.
The user will be sending command from the LabVIEW software and the code will
be deployed on FPGA controller. This will send signals to the stepper modules
inserted in the backplane of cRIO 9074 that in return sends the direction and step
signals to the motors attached with IGUS-Robolink. This allows the motors to move
independently to their respective positions.
3.7 Connecting cRIO 9074 and IGUS-Robolink
For setting up and configuring the cRIO 9074 for this research project, specific
hardware and software were installed. The process for wiring the chassis and
connecting to cRIO is in appendix L. The hardware and software setup required for
this is listed below:
3.7.1 Hardware required
The following hardware was required to be installed:
o Power supply for the controller
o Ethernet connection cable or cross over cable
o C Series stepper drives modules. Datasheet is attached in the appendix G
o 8 Channel, 5V/TTL high speed bidirectional digital input/output module.
Datasheet is attached in the appendix I
Figure3. 7: Complete architecture with NI 9501 (NI, 2011b)
©S.B.Sial, 2013
21
o Power supplies for the connection of robotic arm and drivers
3.7.2 Software required
The following software was required:
o NI LabVIEW 12.0 (or any version 8.6 or later)
o NI LabVIEW Real time module 12.0 (or any version 8.6 or later)
o NI LabVIEW FPGA module 12.0 (or any version form 8.6 or later)
o NI-RIO 12.0 (or any version 8.6 or later)
3.7.3Hardware connection
Once the controller is configured successfully by installing the specific hardware
and software as mentioned above, IGUS-Robolink was connected with NI hardware
then. There were a total of 8 slots available on the chassis, of which 5 were used by
NI 9501 stepper motor drivers, one for each joint of the robotic arm. The pin
configuration for connecting the motors with the drive modules is in appendix K.
All the joints of the robotic arm used in this research were provided with
incremental encoders as discussed earlier, in order to keep track of the position of
the joints. The module used for the connecting this incremental encoder to the
CompactRIO was NI 9401 shown in Fig. 3.8. The datasheet of this module is
attached in the appendix I. This is an 8 channels, 5V/TTL high-speed bidirectional
digital input/output module connected with the encoder wires of IGUS-Robolink.
The whole hardware i.e. NI hardware and IGUS-Robolink is then interfaced with
LabVIEW software in order to develop an algorithm that would be able to move the
robot in a natural manner to express various emotions to the user.
Figure3. 8: NI module 9401 (NI, 2012d)
©S.B.Sial, 2013
22
3.8 Summary of the chapter
This chapter discusses in detail about the hardware and the software platform that is
used for this research. The NI hardware used is cRIO 9074, stepper driver modules
9501 and an encoder module 9401. The reason for choosing this hardware and
connecting this with IGUS-Robolink is described in detail. Method of how to
interface the whole hardware with the software of LabVIEW is also mentioned. It
also highlights the DH parameters used for this specific robot.
The next chapter will discuss the modules and methods used for designing of the
algorithm for this specific platform.
©S.B.Sial, 2013
23
CHAPTER 4
DESGIN OF ALGORITHM USING LABVIEW
4.1 Introduction to design of algorithm
There are basically two main code files that are programmed. One is called the
FPGA VI, which is deployed on the hardware of the cRIO by compiling the code
and generating results. The other VI is the Real Time (RT) VI. This VI serves an
interactive panel for the user to operate the platform.
There are several main modules used for programming of both of these VIs as
named below:
NI LabVIEW FPGA Module
NI LabVIEW Real Time Module
NI LabVIEW SoftMotion Module
NI LabVIEW Robotics Module
These modules will be discussed in detail one by one in this chapter.
4.2 NI LabVIEW FPGA module
The main purpose of using FPGA is to achieve the parallelism in dataflow. NI offers
the FPGA based reconfigurable input/output hardware cRIO to achieve this concept
of parallelism using graphical programming.
The same graphical interface is used for programming real-time as well as FPGA
targets. For this purpose LabVIEW takes its graphical code diagram to different
compilers to create an executable file suitable for specific type of hardware.
LabVIEW offers FPGAs with millions of gates for complex programming with
inherent capacity of parallel programming. The software module used is shown in
Fig. 4.1.
©S.B.Sial, 2013
24
Communication of this FPGA VI with the real time host VI is another important
aspect of FPGA programming. For this LabVIEW provides different host interfaces.
NI offers PC, PXIs, PCIs, and Ethernet enabled hardware that takes care of all the
functions and does not require any custom work by the user. The user can focus on
algorithm design, whereas the hardware takes cares of things like data
communications, direct memory access, registers, bus communication, analog and
digital outputs, clocks, interrupts etc.
The current targets for LabVIEW FPGA include the following hardware shown in
Fig. 4.2. The rugged platform of cRIO is perfect for standalone and network
applications.
Figure4. 1: FPGA programming palette
Figure4. 2: FPGA based hardware offered by NI (Kuhlman, 2013)
©S.B.Sial, 2013
25
Some of the common applications of LabVIEW FPGA systems are as follow
(Kuhlman, 2013.):
o High speed control
o Smart Data acquisition system DAQ
o Digital communication protocols
o Sensor simulation
o On board processing and data reduction
o Co-processing
The block diagram in Fig. 4.3 explains the deployment and creation of bitmap file
for FPGA code:
4.3 NI LabVIEW Real-Time module
The operating system is mainly responsible for managing hardware and hosting
applications on the computer. The real time operating system does these tasks with
high reliability and very precise timings.
NI Real-Time modules give the user the ability to develop a complete and reliable
embedded system that is run by the graphical programming platform of LabVIEW
(NI, 2013c). The real time hardware systems introduced by LabVIEW are NI
CompactRIO, NI Single-Board RIO, PXI, PC and various others.
There are several reasons for using real time system for this project:
o The graphical interface of LabVIEW allows the user to program their tasks
more quickly and easily. The same graphical programming platform is used with
LabVIEW real-time module to create stand-alone systems.
o The common LabVIEW programming system uses the windows operating
system which is not optimized to handle tasks for critical timings over an extended
LabView
FPGA
Xilinx
Compiler Bit file
User generated .VHDL FPGA target
Auto generated
Figure4. 3: Process of code deployment on FPGA
©S.B.Sial, 2013
26
period of time. This module provides the real time operating system for high
reliability and precise timings for the tasks.
o With this module the user can take advantage of various LabVIEW libraries
(e.g. PID, FFT)
4.3.1 Basic Real-Time architecture
Fig. 4.4 shows the basic architecture for real-time module of LabVIEW. The host
program develops the network communication with the target program and then
executes it on the basis of priority given to the loops by user.
4.3.2 Basic Real-Time toolkits
There are several built-in libraries and toolkits available for real-time programming
of LabVIEW that allows the user to concentrate on its logic and make programming
much easier. The Fig. 4.5 shows the toolkit used for real-time programming.
4.3.3 Steps for the development of the Real-time system
The three main steps for developing of the real-time system include (NI, 2013a):
o Development of the application on the host computer i.e. graphical coding for
the system
o Downloading of the code to the real-time hardware target
Figure4. 4: Basic architecture (NI, 2013a)
Figure4. 5: RT toolkit
©S.B.Sial, 2013
27
o Execution of the code
4.4 NI LabVIEW SoftMotion Module
NI has introduced another module called SoftMotion that is compatible with the RT
module and FPGA module. This module helps build custom motion control
applications by providing functions such as path planning, trajectory generation,
position velocity control for various different kinds of steppers as well as servo
motors.
The C-Series driver discussed in the chapters above is used for the motion control of
stepper motors in this research. This module provides various interactive tools for
high level motion functions for simplified development (NI, 2012e).
In this research the main reason for using this module is to create the spline motion
for the stepper motors using the 9501driver. The programming palette for soft
motion module is shown in Fig. 4.7.
Figure4. 6: Summary of development process (NI, 2013e)
Figure4. 7: SoftMotion palette for programming
©S.B.Sial, 2013
28
In order to control the motion of the stepper motors there is a sub-palette called
stepper available inside the motor control. From there the spline motion generation
function is used for the motion control of stepper motors. Fig. 4.8 shows the specific
palette that is used:
The spline function is used to smooth the motion of the motor. This takes data from
spline function and return the step and direction for the operation of motor. Various
types of interpolations are available to generate splines like linear interpolation,
cubic B spline, and Catmull-Rom spline. Cubic B spline interpolation was used in
this research.
Fig. 4.9 shows the stepwise execution of this module. The spline engine and
trajectory generation processes are done on the FPGA side whereas the supervisory
control is done on the RT side.
The user gives commands to the supervisory control loop that is actually the main
loop for motion control. This loop monitors inputs/outputs and faults. This part of
the code is executed on the RT side. This sends commands to the trajectory
generation loop that is actually the path planner. It creates the set points that are used
to calculate the interpolated position by the spline engine which results in smooth
motion. The control loop then creates a command signal based on the set points
USER
COMMAND
SUPERVISORY
CONTROL
TRAJECTORY
GENERATION
Spline
Engine CONTROL
LOOP
Figure4. 8: Spline generation function
Figure4. 9: Spline generation loop
©S.B.Sial, 2013
29
generated by the trajectory generation as shown in Fig. 4.10. The user specifies the
command from user interface that goes to RT side in supervisory control loop. This
sends the signal to FPGA side where trajectory generation process is done and
points are interpolated. This sends the command and direction signal to driver
modules of the stepper motors and operates the robot.
4.5 NI LabVIEW Robotics Module
Another important module used for the programming of this project is the Robotics
module. This module uses software tools to design autonomous and semi-
autonomous systems (NI, 2012f). It includes the following features as shown in Fig.
4.11.
Figure4. 10: Detailed process for Spline generation (NI, 2013f)
Figure4. 11: Robotics toolkit
©S.B.Sial, 2013
30
The main purpose of the robotics module in this research was to use the inverse
kinematics function to generate point-point motion of the robotic arm. The robot
moves automatically to the X, Y, and Z coordinates given by the user provided it is
reachable and within the joint limits of the robotic arm. Fig. 4.12 shows the function
of inverse kinematics in the robotics module.
Fig. 4.13 shows the front panel for setting the parameters for the inverse kinematics.
The parameters shown in the figure are for the specific robot that was used in this
project.
The joints types are set to be revolute as all five of the joints of the robot are
revolute. The twist angle between the first two and last two joints is -1.5radians or
90°. The length of the links for each of them is 4000mm or 0.4m. For a random
point using these parameters, Fig. 4.14 is generated by the inverse kinematics
module.
Figure4. 12: Function of inverse kinematics
Figure4. 13: Parameters for inverse kinematics of robot
©S.B.Sial, 2013
31
This part of code is later integrated in main code that is attached in the appendix J.
Thus, using the inverse kinematics module of LabVIEW it allows the user to input
the spatial coordinates so that the robot moves to the position specified by the user
based on the constraints applied.
4.5.1 Stages involved in implementing inverse kinematics
The code development for the research initiated from building up the logic for
forward kinematics of the robot. The logic was developed so that user will be
entering 5 different angles for each of the motor. The motors will move to the
respective positions as entered by the user independently. However the motion
obtained out of this was a typical trapezoidal motion with sharp edges. The aim was
to smooth down the motion so that the robot moves in a natural way. For this reason
SoftMotion module of LabVIEW was used. However another approach of PID was
also considered before applying the SoftMotion module.
Later for getting on to the inverse kinematics of the robot, the model as shown in
Fig. 6.17 was developed in “Robot simulation model builder” in LabVIEW keeping
in regard the original parameters. However the problem of Direct Memory Access
(DMA) channels was encountered while developing logic for inverse kinematics. As
cRIO 9074 offers only 3 DMA channels for the transfer of data between RT to
FPGA side. To move 5 motors independently, either 5 DMA channels were required
or a logic should be developed that can send data from same DMA channel but
independently for different motors. Different methods that include interleaving and
Figure4. 14: Serial arm with 5 revolute joints
©S.B.Sial, 2013
32
decimating of data, join-split function for data, 2D array approach and clusters etc.
were used. The first two methods failed as they were not moving motors
independently. As soon as the array data of one motors finishes the other motor
stops. However the methods mentioned later were not supported on FPGA side.
Therefore logic was developed so that the smaller array would be sending ‘0’ as its
data elements till the larger array is completely transferred to FPGA side. Thus
making the motors move independently to different positions.
Another major problem encountered was the space on FPGA controller after
developing the correct logic for inverse kinematics. This was resolved by optimizing
area on FPGA before compiling the code and changing the virtual RAM of
computer. Later the Robotics and SoftMotion were integrated in the main code.
4.6 Summary of the chapter
This chapter highlights the modules that were used for the logic development of this
research. It explains about the spline engine generation and the trajectory loop
process that is done on FPGA side. It discusses how the forward and inverse
kinematics for point-point motion was developed and the problems involved in
developing the logic for inverse kinematics. The code for the process is attached in
appendix O.
The next chapter will discuss in detail the core research methodologies that were
used for user’s perception of emotions for specific gestures of the robot. The reasons
for choosing those specific models are also discussed in next chapter.
©S.B.Sial, 2013
33
CHAPTER 5
CORE RESEARCH METHODOLOGY FOR
PERCEPTION OF ROBOT EMOTIONS
5.1 Core concept for research
There are various different emotional models that are introduced by psychologists
for verbal as well as non-verbal communication of machines e.g. Russell’s
circumplex model of affect, PANAS, PANAS-X, PAD scale, SAM scale,
Schimmack and Grob model etc. (Tuomas at el, 2011). Most of these models
interpret the emotions of the machines based on the subjective opinion of people
looking at them.
Affectionate robotic pets, household robots, nursing robots and many other are
creeping in our lives very fast. The future is crowded with emotive humanoid robot
companions (Dautenhahn at el., 2009). Because of this growing importance of
human interaction with robots and the idea of perceiving these robots as social
actors (Dautenhahn K., 1999)., it is important to define this human-robot interaction
in terms of robots working with humans. The focus of this research is how humans
perceive the emotions of robots which look like machines (as opposed to
anthropomorphic robots). It uses three different models to describe the emotional
state of the IGUS robotic platform. These three models will be discussed later in this
chapter. Participants were invited to observe three different gestures of the robot in
three states, and afterwards they had to fill in a questionnaire that would tell how
they have perceived the emotional state of robot for a particular gesture at that time.
The perception of emotions by the participants greatly depends on the type of
embodiment used in the research. According to Mori’s Uncanny Valley theory the
degree of expectation towards the robot increases as it becomes more human-
looking. However in his theory there is a point on the anthropomorphic scale where
the robot’s appearance becomes confusing and it is difficult to distinguish between
humans and robots (Mori, 2005). This was proved in the research carried by
Bartneck et al. in which all the robots were not treated as the same by the
©S.B.Sial, 2013
34
participants because of their anthropomorphic or zoomorphic appearance which
differed from each other (Bartneck et al., 2006).
In this particular research the robot is not from the anthropomorphic or zoomorphic
family of robots. It looks like a machine, so the expectations from the participants
are lowered because of the type of embodiment. Moreover as the robot cannot
exhibit as many moods as an anthropomorphic or zoomorphic robot can, the moods
are measured on wide zones on the standard scales used for this research.
5.2 Modelling machine emotions
Emotions in machines are very important in today’s world of research. People may
use an emotionless machine as a tool but not a reliable partner or companion to work
with them in industry and various other fields. So emotions are important aspect in
human-machine communication. Although even with high-level programming
techniques it will be difficult for the machines to express emotions like humans do.
According to research even simple machines can express or can display emotional
feelings (Braitenberg, 1984). The example of two vehicles controlled by sensors as
shown in Fig. 5.1 clearly explain the concept.
Both of these are connected with the sensors that respond to a light source. When
these vehicles are exposed to light, the one on right moves towards the light source
and the left away from the source. This difference is because of the way sensors are
connected to motors. The left cart will move away as its right sensor is receiving
stronger light then left, thus producing more torque on right wheel than left. This
motion can be translated in terms of emotional feeling that as the right vehicle does
not like the light source so it is moving away; whereas the other cart is attracted
Figure5. 1: Emotional vehicles responding to light (Nishida et al., 2010)
©S.B.Sial, 2013
35
towards the light source (Nishida et al., 2010). The computational model for these
emotional vehicles is shown in Fig. 5.2
5.3 Emotional models for research
There are many different models for categorising the emotions of robot by the user.
Detailed study for mutual co-relation among different kinds of emotions is being
done. The research is carried out by using different statistical approaches, for
example, multidimensional scaling, and factor analysis of individual reports of
different emotional experiences. The research consistently resulted in 2-D models of
affective emotional experiences, with different attributes for the two dimensions
such as valence and arousal by Russell, dimensions of PANAS by Watson, tension
and energy by Thayer and various others (Posner et al., 2005). The ones that are
used in this research are:
o Russell’s circumplex Model
o Tellegen-Watson-Clark model
o PAD scale
The reason for selecting these models is because these are well-known and
renowned one’s that are used for research of HRI. The models will be discussed and
explained one by one.
5.3.1 Russell’s circumplex model of affect
There are two basic models used for measuring the emotional state of machines that
have found wide acceptance and support. These are Ekman’s and Russell’s model.
Russell’s model is used for this research as from the Ekman’s model “it is not clear
Figure5. 2: Computational model for emotional vehicle (Nishida et al., 2010)
©S.B.Sial, 2013
36
which emotions make the basic set of which all other emotions can be constructed”
(Saerbeck and Bartneck, 2010).
Russell proposed the basic circumplex model of emotions (Russell, 1980). This
model described emotion in two axis space. The vertical axis represents the arousal
in the observed emotion and the horizontal axis represents valence. The center where
both axes meet was neutral emotions. This model was usually used for testing
stimuli of emotion words, emotional facial expressions and affective states
(Remington et al., 2000).
Most psychologists believe that emotions are independent from each other and have
their own dimensions such as distress, depression and anxiety etc. However Russell
proposed that all these affective emotional states are interlinked and dependent on
each other (Russell, 1980). He proposed a circular model in a two dimensional bi-
polar space of valence and arousal rather than as a mono-polar space that are
independent of each other. Later, the model was extended for 28 different feelings
that are interlinked and sometimes synonymous. The particular Russell’s model
used in this research is shown in Fig. 5.3:
Because there are no distinct boundaries between emotions like “happiness” and
“being pleased”, “relaxed” and “clam”, “sad” and “gloomy” etc. emotions that
overlap each other are placed close together making a cluster in this scale (Russell,
1980). Fig. 7.3 shows 28 emotional states distributed in four different quadrants
following core concept of two main axes of valence and arousal. As defined by
Figure5. 3: Russell’s circumplex model of emotions (Russell, 1980)
©S.B.Sial, 2013
37
(Kensinger, 2004) “The dimension of valence ranges from highly positive to highly
negative whereas the dimension of arousal ranges from calming or soothing to
exciting and agitating.” The valence axis can be defined in terms of unpleasant-
pleasant axis and the arousal can be defined as deactivation-activation in the
emotions (Junghyun et al., 2010). Thus there can be various different events that can
be negative and agitating or positive, calming and soothing. In this scale the Table
5.1 represents the emotions with respect to degrees in this circular arrangement.
All these emotions are placed on the circular pattern graph keeping in view the
relation of these with arousal and valence. For example “delighted” is placed at
24.9° indicating that it has both factors of arousal as well as pleasure (Russell et al.,
1989). Similarly, looking at the emotion of being “excited” at 48.6° involves higher
arousal. Furthermore, looking at the behaviour of being “astonished” we can see less
pleasure and more arousal (Russell et al., 1989). Words that are close to each other
on the graph describe similar emotions, whereas being apart on scale and further
from each other indicates the difference in emotional states. This scale with 28
different emotions in two bipolar spaces was used in the questionnaire for
interpreting the emotional state of the robotic platform.
There were two other more or less similar scales that were proposed by Russell, but
experimentation for quantitative comparison among the scales showed that these
produce equivalent results. All of the scales look almost the same and produce
similar results (Russell, 1980). The one used in this research is shown in Fig. 5.3
EMOTIONS DEGREES
Pleasure 0°
Excitement 45°
Arousal 90°
Distress 135°
Misery 180°
Depression 225°
Sleepiness 270°
Contentment 315°
Table5. 1: Location of emotion on circular graph (Russell, 1980)
©S.B.Sial, 2013
38
The reason for not using the basic model is that it “accounts for the substantial
proportion, but not all, of the variance in self-reported affective state” (Russell,
1980).
5.3.1.1 Example of affective interaction
There is a SMS-service called eMOTO that sends text messages in addition to
different colourful and animated shapes in the background that express the emotions
of the person sending text. The expression is chosen on the basis of set of gestures
using stylus pen that comes with sensors that would know about the pressure and the
shakiness in movements. Thus the background would be representing the emotions
of person sending the text. It also allows the user to build their own gestures as they
are not limited to specific set of gestures only. Pressure and shaking movements is
the main constituent for expressing these emotions (Höök, 2013).
Fig. 5.4 (a) shows various physical movements that have different pressure and
shakiness in them. These can be related with the affective experiences of Russell’s
circumplex model of affect shown in Fig. 5.4 (b). These emotions are then mapped
to colourful expressions in Fig. 5.4 (c).
5.3.2 Tellegen-Watson-Clark model
Another model used for the analysis of emotional moods by perceiving the motion
of the robot is Tellegen-Watson-Clark model. This model is shown in the Fig. 5.5:
(a) (c) (b)
Figure5. 4: Russell’s circumplex model for affective interaction (Höök, 2013)
©S.B.Sial, 2013
39
This model is another way of rating the emotions that eventually emerged as
prominent criteria. The two main dimensional ratings in this scale are PA and NA.
PA is an abbreviation for positive affect and is the degree of positive emotions that
are being felt like “being cheerful” and “enthusiastic” etc. Whereas the Negative
affect (NA), is the extent of experiencing negative moods like anger, rage, guilt etc.
(Coan and Allen, 2007).
The wide scope of these terms includes many emotions in them. PA includes all
those feelings that are pleasant like being enthusiastic, confident, interested, healthy
etc. High positive effects reflect the state of full energy and full concentration
whereas low PA is taken to be in the state of calmness and serenity.
On the other hand, NA includes all the negative feeling and emotions and moods of
being guilty, feeling angry, fearful, distressed etc. Low NA is to be sad and being in
a lethargic mood etc. However according to (Yang and Lee, 2004) NA is difficult to
distinguish from each other as they are very closely related unlike PA. An example
of anger and guilt is explained in his paper. Both of these are highly NA and are
placed together, however taking the emotions of sadness and guilt, they are
separated from each other, as sadness is also towards PA. Although PA and NA is
highly uncorrelated and it is easy to distinguish between these two emotions yet to
distinguish among the emotions in each category is difficult. “Happiness and
Figure5. 5: Tellegen-Watson-Clark model (Trohidis et al., 2011)
©S.B.Sial, 2013
40
sadness form a largely unidimensional bipolar structure, but PA and NA are
relatively independent” (Tellegen et al., 1999).
This model represents the emotional state with more clarity from another new
perspective of PA and NA as two independent emotional axes. Mainly two pairs of
dimensions are being mapped by this model. Firstly PA and NA and the other
dimension rotated to 45° showing the emotions of pleasantness VS unpleasantness.
Including this dimension, give rises to the circular shape of model.
There was a great deal of research addressing the question whether these PA and NA
are independent or not. But research has shown that these have emerged as two
independent and consistent scales for categorising emotions (Watson et al., 1988).
However this model was renamed to avoid the ambiguity of terminology later on.
PA and NA were later called Positive Activation and Negative Activation (Tellegen
et al., 1999). Watson et al. concludes on the basis of their research that PA and NA
are reliable and efficient ways of measuring two important dimensions of moods
(Watson et al., 1988).
5.3.3 PAD scale :
The third scale used for measuring the emotional state is PAD shown in Fig. 5.6.
This scale was developed by Albert Mehrabian to describe the different states of
emotions in terms of Pleasure, Arousal and Dominance as three independent
orthogonal features. It uses these three independent dimensions to describe all the
emotional states (Mehrabian, 1980). This scale is used for non-verbal
communication such as body language etc., in psychology (Mehrabian, 1977).
There are three main dimensions for this scale in terms of which emotions are
measured:
o The Pleasure-Displeasure scale: This particular dimension of the scale
measures how pleasant or unpleasant an emotion is. For example happy and
excited both comes under the category of pleasant emotions. However anger
comes under displeasure.
o The Arousal-Nonarousal scale: This is another independent dimension of
the PAD scale to measure the Arousal or Nonarousal aspect in the emotions.
©S.B.Sial, 2013
41
It basically measures the intensity of emotions. The emotions that fall in the
same category of either being pleasant or unpleasant can be further
categorized on the basis of their intensity. For example, talking about two
unpleasant emotions of anger and rage, the intensity for both of them is
different although both come under the feeling of unpleasantness. Anger has
less arousal than that of rage. Similarly, being happy and excited is another
example from the pleasant category of emotions. Though both fall under
same index but being happy contains less amount of arousal than that of
being excited.
o The Dominance-Submissiveness scale: The third dimension of the PAD
scale measures the factor of being dominant or submissive in its emotions.
For example taking two pleasant emotions of happy and excited. Excitement
contains more dominance. Similarly anger is more dominant than fear
although both are under the list of unpleasant feelings.
For this research these three independent parameters were used to evaluate the
emotions. A table was created and for each motion of the robotic arm, people were
asked to mark all of the three factors of Pleasure, Arousal and Dominance in terms
of being high, low or medium and then marking the overall affect they are getting
from the motion. Before start filling in the questionnaire, people were clearly
explained what these scales mean.
These three independent orthogonal dimensions of the PAD scale provide detailed
information about the emotional state. The foundations for this PAD scales involve
the differentiation between the emotions and temperament (Mehrabian, 1996).
According to this scale any point in the space of PAD scale represents the emotion.
Rather than measuring it in terms of values ranging between 0-1, this research
adopts another method of marking the emotional state in terms of High, Medium and
Low states of Pleasure, Arousal and Dominance. This is because there are only three
emotional states that are tested for three different gestures. Moreover it becomes
easy for the participant to mark it in this way. For the validity of this scale, it is
compared with the results of other two scales.
©S.B.Sial, 2013
42
People were also asked to judge the overall effect for the perceived emotion. The
overall affect is categorised in three main groups of sad/tired---unpleasant,
happy/pleased---pleasant, and excited---aroused. The overall emotion is then
compared with the individual factors marked by the participants.
5.4 Selection of gestures
For all of the scales there were three different gestures. The emotions for all these
three gestures were measured on these scales by dividing the quadrants according to
the emotional state. The three gestures selected were:
o Point-Point motion
o Waving of the robotic arm
o Bowing down to welcome
5.4.1 Graphical illustration of gestures :
The point-point motion is the most basic and general kind of movement in the world
of robotics. Fig. 5.7 represents point-point motion of robot from home position to
three different points. This is done using the concept of inverse kinematics. The
robot moves in a smooth spline trajectory as explained in earlier chapters.
D (Dominance)
P (pleasure)
A (Arousal)
Figure5. 6: PAD model
©S.B.Sial, 2013
43
Fig. 5.8 represents the graphical illustration of waving of robot in a clockwise
pattern. This gesture was selected as it is a basic human gesture in the form of a
repetitive movement.
Fig. 5.9 represents the graphical illustration of bowing down of the robot form right
to left. The reason for choosing this gesture was as it is a universal cultural gesture
that people can recognize quickly and easily. The robot can be seen bowing from
straight position to almost 90 degrees for its last joint in Fig. 7.10.
Figure5. 7: Graphical representation of point-point motion
Figure5. 8: Graphical representation of waving of robot
©S.B.Sial, 2013
44
5.5 Motion characteristics
The motion characteristics that were used to change the emotional state of the robot
were velocity and acceleration. Because of the change of these two parameters the
robot changed its speed, trajectory, time consumed and curvatures (Saerbeck and
Bartneck, 2010). The changing values of the velocity and acceleration shows the
prominent change in emotional effect perceived by the user.
The effect can also be observed in the spline motion graph that is generated in the
LabVIEW code.
However there were physical constraints when choosing the values such as, if the
values were too high the wires could be pulled off the drive wheel. Similarly, if the
velocity and acceleration were too low slippage occurred and the motor did not
rotate properly.
Fig. 5.10 to 5.15 shows that as the motion parameters are changed spline shown on
the graph is also changed. The splines for Point-Point motion of robotic arm that is
for Gesture1 for all three parameters are shown in Fig. 5.10 to 5.12.
Figure5. 9: Graphical representation of bowing of robot
©S.B.Sial, 2013
45
Gesture 1: At V=250, A=10
Gesture 1: At V=800, A=50
Gesture 1: At V=2000, A=300
Figure5. 10: Spline curve for G1 at V=250
Figure5. 11: Spline curve for G1 at V=800
Figure5. 12: Spline curve for G1 at V=2000
©S.B.Sial, 2013
46
The spline curves for Gesture 2 that is waving of the robotic arm, considering all the
three set of parameters are shown in Fig. 5.13 to 5.15:
Gesture 2: At V=100, A=15
Gesture 2: At V=100, A=5:
Gesture 2: At V=100, A=1.5:
Figure5. 13: Spline curve for G2 at A=15
Figure5. 14: Spline curve for G2 at A=5
Figure5. 15: Spline curve for G2 at A=1.5
©S.B.Sial, 2013
47
5.6 Summary of the chapter
This chapter explains the emotional models and reasons for choosing the specific
models. It focuses on the fact of how three different emotional models i.e. Russell’s
circumplex model of affect, Tellegen-Watson-Clark model and PAD scale can be
used by the user to mark the perceived emotion of a mechanoid robot. The gestures
and motion parameters chosen to change these gestures are also explained in detail.
The next chapter discusses the experiments performed using these emotional models
and gestures and the results collected. The discussion on how these models give the
user flexibility of marking the perceived emotions for particular gestures in certain
set of range are covered in next chapter.
©S.B.Sial, 2013
48
CHAPTER 6
EXPERIMENTS AND RESULTS
6.1 Experiments for emotional communication
Several experiments were designed to investigate the emotional communication of
the robot. These experiments were based on the emotional scales discussed above.
The robot exhibited a variety of gestures, with changing motion characteristics and
the participants had to judge the emotional state by marking the questionnaire given
to them. The ethical approval and the consent form for this research is attached in
appendix N.
6.1.1 Experiment procedure
Fig. 8.1, 8.2 and 8.3 shows the participant sheet that they were given in the
experiment. The complete data was collected from 18 participants including males
and females that fall in the age group of 17-50 years. The experiment took
approximately 20 minutes.
Each session was started with a brief introduction of the project and an explanation
of how the participant would have to mark the perceived emotions on the given
scale. Participants were told what the individual terms mean. After this introductory
session, the participants were given the consent form to sign before they start
observing the robot.
In total 3 gestures, each with 3 different sets of velocity and acceleration were
shown to each of the participants. This setup resulted in 3x3 emotions marked
independently on each of the scales. So in total of 3x (3x3) emotion were marked by
each of the participants for all of the three scales i.e. three models each with three
different gestures, each with three different subsets of velocity and acceleration. For
Russell’s model, arousal and valence were the two independent axes. For the
Tellegen-Watson-Clark model PA and NA were the main independent parameters
and for the PAD scale Pleasure, Arousal and Dominance were used as a set of
independent parameters to measure the emotions reflected in the motion of robot.
Figure 70: PAD model [70]
©S.B.Sial, 2013
49
To have some reference for the comparison of different motions, the participants
were shown the specific gesture for all the three different values of velocity and
acceleration and then were asked to mark the perceived emotions for each set of
velocity and acceleration. If requested, the participants were shown the motion with
specific parameters again.
Participants marked a circle on the specific emotion that was perceived for that
motion on the model graph. Some of the participants marked more than one emotion
for the same motion. However they were in the same quadrant and closely
resembled each other. In other words they can be termed as overlapping emotions.
6.2 Questionnaires for measuring perception of emotions
There were three different questionnaires based on the different scales that each of
the participants had to fill in. The questionnaires for three different scales and
gestures presented to the participants are shown in Fig. 6.1 to 6.3.
Fig. 6.4 to 6.6 shows the sample of filled questionnaire by the same participant for
Russell’s model, Tellegen-Watson-Clark model and PAD scale respectively at
different velocities and accelerations for three different gestures.
The units used for velocity and acceleration are counts/revolutions and
counts/revolutions^2 respectively for all three gestures and scales.
©S.B.Sial, 2013
50
6.2.1Questionnaires for Russell’s model
Russell’s model of mood QUESTIONNAIRE
GESTURE: 1 POINT –POINT MOTION:
VELOCITY (Counts/rev)
ACCELARATION (Counts/rev^2)
REPRESENTATI-ON ON GRAPH
30 30 3A
50 50 3B
100 100 3C
VELOCITY
(Counts/rev)
ACCELARATION
(Counts/rev^2)
REPRESENTATION ON GRAPH
250 10 1A
800 50 1B
2000 300 1C
VELOCITY
(Counts/rev)
ACCELARATION
(Counts/rev^2)
REPRESENTATION ON GRAPH
100 15 2A
100 5 2B
100 1.5 2C
GESTURE: 2 WAVING OF ROBOTIC ARM:
o GESTURE: 3 BOWING TO WELCOME:
:
I
III
II
IV
I
III
II
IV
I
III
II
IV
Figure6. 1: Russell's model questionnaire
©S.B.Sial, 2013
51
6.2.2 Questionnaires for Tellegen-Watson-Clark model
Tellegen-Watson-Clark model of mood QUESTIONNAIRE
GESTURE: 1 POINT –POINT MOTION:
GESTURE: 2 WAVING OF
ROBOTIC ARM:
GESTURE: 2 WAVING OF ROBOTIC ARM
GESTURE: 3 BOWING TO WELCOME:
VELOCITY (Counts/rev)
ACCELARATION (Counts/rev^2)
REPRESENTATI-ON ON GRAPH
250 10 1A
800 50 1B
2000 300 1C
VELOCITY (Counts/rev)
ACCELARATION (Counts/rev^2)
REPRESENTATI-ON ON GRAPH
100 15 2A
100 5 2B
100 1.5 2C
VELOCITY (Counts/rev)
ACCELARATION (Counts/rev^2)
REPRESENTATI-ON ON GRAPH
30 30 3A
50 50 3B
100 100 3C
I
III
II
IV
I
III
II
IV
I
III
II
IV
Figure6. 2: Tellegen-Watson-Clark model questionnaire
©S.B.Sial, 2013
52
6.2.3 Questionnaires for PAD model
PAD QUESTIONNAIRE
Figure6. 3: PAD questionnaire
©S.B.Sial, 2013
53
6.2.4 Measurement of emotions by participants
Filled sample questionnaire for Russell’s circumplex model of affect
Figure6. 4: Russell's questionnaire filled by the participant
©S.B.Sial, 2013
54
Here is an example of the questionnaire filled by the same participant for the
Tellegen-Watson-Clark model at different velocities and accelerations for three
different gestures:
Figure6. 5: Tellegen-Watson-Clark questionnaire filled by the participant
©S.B.Sial, 2013
55
Here is an example of the questionnaire filled by the same participant for the PAD
model at different velocities and accelerations for three different gestures:
Figure6. 6: PAD questionnaire filled by the participant
©S.B.Sial, 2013
56
6.3 Emotion recognition based on scales The emotions for each of the gestures were marked by the participants based on the
perceived affect. The first two scales were divided into four different quadrants
based on the range of emotions. So, the participants marked the emotions quadrant-
wise for each of the gestures for both Russell’s and Tellegen-Watson-Clark models
of emotions.
Emotional ranges for each quadrant for both scales are given below:
For Russell’s model
Q1: Excited---Aroused
Q2: Tense---Annoyed---Miserable
Q3: Tired---Sad---Miserable
Q4: Calm---Content---Pleased
For Tellegen-Watson-Clark model
Q1: Alert/Delighted--- Amazed/Surprized
Q2: Pleasant---Happy/Joyful
Q3: Sleepy---Calm/Relaxed
Q4: Unpleasant---Sad/Tired
The third scale that is PAD is actually based on measuring the pleasure, arousal and
dominance in the emotions. Thus this scale cannot be divided into quadrants. So
another approach of measuring the overall effect was implemented. The participants
were asked to mark from the range of three different overall effects of the emotions
that they perceived from the motion of embodiment after marking the three
individual factors of pleasure, arousal and dominance. The overall emotional range
for this scale used is given below:
For PAD model
1. Sad/Tired---Unpleasant
2. Happy/Pleased---Pleasant
3. Excited/Aroused---Alert
This overall result is then compared with the individual effects marked on the scale
that is shown in the results.
©S.B.Sial, 2013
57
6.4 Models results
6.4.1 Results for Russell’s model
The graph in Fig. 6.7 shows that at low velocity and acceleration 17/18 people have
marked it in Q3 shown by green bar that says the perceived emotion is Tired/Sleepy
--- Sad/Miserable. For medium level of velocity and acceleration 7/18 participants
marked it in Q4 shown by purple bar saying that the robot is Calm/Content ---
Pleased and at high velocity and acceleration majority i.e. 12/18 people have marked
it in Q1 i.e. Excited/Delighted ---- Aroused.
Gesture:1 Q1 Q2 Q3 Q4
V=250 & A=10 0 0 17 1
V=800 & A=50 4 6 1 7
V=2000 & A=300 12 5 0 1
Q1: Excited/Delighted ---- Aroused
Q2: Tense/Annoyed --- Miserable
Q3: Tired/Sleepy --- Sad/Miserable
Q4: Calm/Content --- Pleased
Table6. 1: Response of participants for Russell’s model G1
0123456789
1011121314151617181920
V=250 & A=10 V=800 & A=50 V=2000 & A=300
Q1
Q2
Q3
Q4
Gesture 1: point-point motion
Figure6. 7: Russell’s model graph for 3 sets of parameters for G1
©S.B.Sial, 2013
58
The graph in Fig. 6.8 shows that 15/18 people have marked the gesture of waving in
category Q3 represented by green bar, falling under the emotions of Tired/Sleepy ---
Sad/Miserable. At medium level of velocity and acceleration 7/18 people marked it
as Calm/Content --- Pleased shown by purple bar. For highest values of velocity and
acceleration majority i.e. 13/18 perceived it as Excited/Delighted ---- Aroused
shown by blue bar.
Gesture:2 Q1 Q2 Q3 Q4
V=100 & A=15 0 2 15 1
V=100 & A=5 1 6 4 7
V=100 & A=1.5 13 5 0 0
Q1: Excited/Delighted ---- Aroused
Q2: Tense/Annoyed --- Miserable
Q3: Tired/Sleepy --- Sad/Miserable
Q4: Calm/Content --- Pleased
Table6. 2: Response of participants for Russell’s model G2
0123456789
1011121314151617181920
V=100 & A=15 V=100 & A=5 V=100 & A=1.5
Q1
Q2
Q3
Q4
Gesture 2: Waving of robotic arm
Figure6. 8: Russell’s model graph for 3 sets of parameters for G2
©S.B.Sial, 2013
59
0123456789
1011121314151617181920
V=30 & A=30 V=50 & A=50 V=100 & A=100
Q1
Q2
Q3
Q4
Gesture 3: Bowing down of robot
The graph in Fig. 6.9 shows that 15/18 people have marked the gesture of waving in
category Q3 represented by green bar, falling under the emotions of Tired/Sleepy ---
Sad/Miserable. At medium level of velocity and acceleration 9/18 people marked it
as Calm/Content --- Pleased shown by purple bar. For highest values of velocity and
acceleration majority i.e. 11/18 perceived it as Excited/Delighted ---- Aroused
shown by blue bar.
Gesture:3 Q1 Q2 Q3 Q4
V=30 & A=30 0 1 15 2
V=50 & A=50 2 3 4 9
V=100 & A=100 11 2 0 5
Q1: Excited/Delighted ---- Aroused
Q2: Tense/Annoyed --- Miserable
Q3: Tired/Sleepy --- Sad/Miserable
Q4: Calm/Content --- Pleased
Table6. 3: Response of participants for Russell’s model G3
Figure6. 9: Russell’s model graph for 3 sets of parameters for G3
©S.B.Sial, 2013
60
6.4.2 Results for Tellegen-Watson-Clark model
The graph in Fig. 6.10 shows that for point-point motion of robot according to
Tellegen-Watson-Clark model 15/18 people have marked the gesture of waving in
category Q4 represented by purple bar, falling under the emotions of Unpleasant ---
Sad/Tired. At medium level of velocity and acceleration 11/18 people marked it as
Pleasant --- happy/Joyful shown by red bar. For highest values of velocity and
acceleration majority i.e. 9/18 perceived it as Alert/Delighted----Amazed/Surprized
shown by blue bar.
Gesture:1 Q1 Q2 Q3 Q4
V=250 & A=10 0 1 2 15
V=800 & A=50 2 11 1 4
V=2000 & A=300 9 1 6 2
Q1: Alert/Delighted---- Amazed/Surprized
Q2: Pleasant --- happy/Joyful
Q3: Sleepy --- Calm/Relaxed
Q4: Unpleasant --- Sad/Tired
Table6. 4: Response of participants for Tellegen-Watson-Clark model G1
0123456789
1011121314151617181920
V=250 & A=10 V=800 & A=50 V=2000 & A=300
Q1
Q2
Q3
Q4
Figure6. 10: Tellegen-Watson-Clark model graph for 3 sets of parameters for G1
©S.B.Sial, 2013
61
The graph in Fig. 6.11 shows that for waving of robot according to Tellegen-
Watson-Clark model 15/18 people have marked the gesture of waving in category
Q4 represented by purple bar, falling under the emotions of Unpleasant ---
Sad/Tired. At medium level of velocity and acceleration 11/18 people marked it as
Pleasant --- happy/Joyful shown by red bar. For highest values of velocity and
acceleration majority i.e. 12/18 perceived it as Alert/Delighted----Amazed/Surprized
shown by blue bar.
Gesture:2 Q1 Q2 Q3 Q4
V=100 & A=15 1 0 2 15
V=100 & A=5 5 11 2 0
V=100 & A=1.5 12 6 0 0
Q1: Alert/Delighted---- Amazed/Surprized
Q2: Pleasant --- happy/Joyful
Q3: Sleepy --- Calm/Relaxed
Q4: Unpleasant --- Sad/Tired
Table6. 5: Response of participants for Tellegen-Watson-Clark model G2
0123456789
1011121314151617181920
V=100 & A=15 V=100 & A=5 V=100 & A=1.5
Q1
Q2
Q3
Q4
Gesture 2: Waving of the robot
Figure6. 11: Tellegen-Watson-Clark model graph for 3 sets of parameters for G2
©S.B.Sial, 2013
62
0123456789
1011121314151617181920
V=30 & A=30 V=50 & A=50 V=100 & A=100
Q1
Q2
Q3
Q4
Gesture 3: Bowing down of robot
The graph in Fig. 6.12 shows that for bowing down of robot according to Tellegen-
Watson-Clark model 13/18 people have marked the gesture of bowing in category
Q4 represented by purple bar, falling under the emotions of Unpleasant ---Sad/Tired.
At medium level of velocity and acceleration 14/18 people marked it as Pleasant ---
happy/Joyful shown by red bar. For highest values of velocity and acceleration
majority i.e. 11/18 perceived it as Alert/Delighted----Amazed/Surprized shown by
blue bar.
Gesture:3 Q1 Q2 Q3 Q4
V=30 & A=30 1 2 2 13
V=50 & A=50 2 14 1 1
V=100 & A=100 11 7 0 0
Q1: Alert/Delighted---- Amazed/Surprized
Q2: Pleasant --- happy/Joyful
Q3: Sleepy --- Calm/Relaxed
Q4: Unpleasant --- Sad/Tired
Table6. 6: Response of participants for Tellegen-Watson-Clark model G3
Figure6. 12: Tellegen-Watson-Clark model graph for 3 sets of parameters for G3
©S.B.Sial, 2013
63
6.4.3 Results for PAD model
Table6. 7: Response of participants for PAD model G1
Gesture: Point-point motion
Pleasure Arousal Dominance
V=250 & A=10 LOW 17 17 14
MED 1 0 4
HIGH 0 1 0
overall
Range 1: R1 sad/tired-unpleasant 18
Range 2: R2 happy/pleased-pleasant 0
Range 3: R3 excited-aroused 0
V=800 & A=50 LOW 4 3 2
MED 13 14 15
HIGH 1 1 1
Overall
Range 1: R1 sad/tired-unpleasant 3
Range 2: R2 happy/pleased-pleasant 11
Range 3: R3 excited-aroused 4
V=2000 & A=300 LOW 4 3 2
MED 4 5 3
HIGH 10 10 13
overall
Range 1: R1 sad/tired-unpleasant 1
Range 2: R2 happy/pleased-pleasant 12
Range 3: R3 excited-aroused 5
0123456789
1011121314151617181920
LOW
MED
HIG
H R1
R2
R3
LOW
MED
HIG
H R1
R2
R3
LOW
MED
HIG
H R1
R2
R3
V=250 & A=10 V=800 & A=50 V=2000 & A=300
P
A
D
overall
Gesture 1: Point-point motion
Figure6. 13: PAD model graph for 3 sets of parameters for G1
©S.B.Sial, 2013
64
The graph in Fig. 6.13 shows that for point-point motion of robot according to PAD
scale 18/18 people have marked the gesture in category R1 represented by purple
bar, falling under the emotions of sad/tired-unpleasant. At medium level of velocity
and acceleration 11/18 people marked it in R2 as Pleasant --- happy/Joyful shown by
purple bar. For highest values of velocity and acceleration majority i.e. 12/18
perceived it as Pleasant --- happy/Joyful shown by purple bar.
Gesture2: Waving of robot
Pleasure Arousal Dominance
V=100 & A=15 LOW
17 15 15
MED
1 1 3
HIGH
0 2 0
overall
Range 1: R1 sad/tired-unpleasant 17
Range 2: R2 happy/pleased-pleasant 1
Range3 :R3 excited-aroused 0
V=100 & A=5 LOW
2 3 4
MED
16 14 12
HIGH
0 1 2
overall
Range 1: R1 sad/tired-unpleasant 3
Range 2: R2 happy/pleased-pleasant 14
Range3 :R3 excited-aroused 1
V=100 & A=1.5 LOW
3 2 0
MED
5 6 5
HIGH
10 10 13
overall
Range 1: R1 sad/tired-unpleasant 0
Range 2: R2 happy/pleased-pleasant 4
Range3 :R3 excited-aroused 14
Table6. 8: Response of participants for PAD model G2
©S.B.Sial, 2013
65
The graph in Fig. 6.14 shows that for waving of robot according to PAD scale 17/18
people have marked the gesture in category R1 represented by red bar, falling under
the emotions of sad/tired-unpleasant. At medium level of velocity and acceleration
14/18 people marked it in R2 as Pleasant --- happy/Joyful shown by red bar. For
highest values of velocity and acceleration majority i.e. 14/18 perceived it as
excited-aroused shown by red bar.
0123456789
1011121314151617181920
LOW
MED
HIG
H R1
R2
R3
LOW
MED
HIG
H R1
R2
R3
LOW
MED
HIG
H R1
R2
R3
V=100 & A=15 V=100 & A=5 V=100 & A=1.5
P
A
D
overall
Gesture 2: Waving of robot
Figure6. 14: PAD model graph for 3 sets of parameters for G2
©S.B.Sial, 2013
66
Gesture3: Bowing down
of robot Pleasure Arousal Dominance
V=30 & A=30 LOW
16 16 16
MED
2 2 2
HIGH
0 0 0
overall
Range 1: R1 sad/tired-unpleasant 18
Range 2: R2 happy/pleased-pleasant 0
Range3 :R3 excited-aroused 0
V=50 & A=50 LOW
2 14 2
MED
4 14 0
HIGH
7 11 0
overall
Range 1: R1 sad/tired-unpleasant 3
Range 2: R2 happy/pleased-pleasant 15
Range3 :R3 excited-aroused 0
V=100 &
A=100 LOW
1 1 2
MED
8 3 5
HIGH
10 14 11
overall
Range 1: R1 sad/tired-unpleasant 1
Range 2: R2 happy/pleased-pleasant 6
Range3 :R3 excited-aroused 11
Table6. 9: Response of participants for PAD model G3
Table8. 1:
0123456789
1011121314151617181920
LOW
MED
HIG
H R1
R2
R3
LOW
MED
HIG
H R1
R2
R3
LOW
MED
HIG
H R1
R2
R3
V=30 & A=30 V=50 & A=50 V=100 & A=100
P
A
D
overall
Gesture 3: Bowing down of robot
Figure6. 15: PAD model graph for 3 sets of parameters for G3
©S.B.Sial, 2013
67
The graph in Fig. 6.15 shows that for bowing of robot according to PAD scale 18/18
people have marked the gesture in category R1 represented by orange bar, falling
under the emotions of sad/tired-unpleasant. At medium level of velocity and
acceleration 15/18 people marked it in R2 as Pleasant --- happy/Joyful shown by
orange bar. For highest values of velocity and acceleration majority i.e. 11/18
perceived it as excited-aroused shown by orange bar.
6.5 Discussion of results
The statistical results concluded are shown below for all the gestures at all set of
velocities and acceleration for the three models. The percentage represents the
number of people that have marked particular emotion in that specific set of
quadrant.
Gesture1: Point-point motion at V=250 and A=10:
For Gesture1: Point-point motion at V=800 and A=50:
For Gesture1: Point-point motion at V=2000 and A=300:
Russell’s model Tellegen-Watson-Clark model PAD model
Q1 0%
Q2 0%
Q3 95%
Q4 5%
Q1 0%
Q2 5%
Q3 11%
Q4 83%
R1 100%
R2 0%
R3 0%
Russell’s model Tellegen-Watson-Clark model PAD model
Q1 22%
Q2 33%
Q3 6%
Q4 39%
Q1 11%
Q2 61%
Q3 5%
Q4 22%
R1 17%
R2 61%
R3 22%
Russell’s model Tellegen-Watson-Clark model PAD model
Q1 67%
Q2 27%
Q3 0%
Q4 6%
Q1 50%
Q2 5%
Q3 33%
Q4 11%
R1 5%
R2 67%
R3 28%
©S.B.Sial, 2013
68
Considering the first parameter for gesture1 of all the three models that is at V=250
and A=10, the result shows that this particular motion of the robot in terms of
emotions perceived by majority of the participants comes under the following
categories for all the models. The percentage shown represents the highest number
of participants that marked in the particular quadrant.
Gesture1: Point-point motion at V=250 and A=10
Russell’s model : Q3-95%: Tired/Sleepy --- Sad/Miserable
Tellegen-Watson-Clark model: Q4-83%: Unpleasant --- Sad/Tired
PAD model: Range1-100%: Sad/Tired-Unpleasant
The results from all the three models come under the same category of being
unpleasant, sad, tired etc. For the PAD model, most of the people marked pleasure,
arousal and dominance as “low” for this set of parameters. This also makes sense
that if the emotions shown by the embodiment are sad and unpleasant then all of the
three factors will fall in low category. The percentage of people that marked in the
categories is shown. The remaining percentage of the people is divided among the
rest of categories for the gesture. Now considering the second set of parameters for
all the three models of gesture1 that is point-point motion at V=800 and A=50, the
following results are obtained:
Gesture1: Point-point motion at V=800 and A=50
Russell’s model : Q4-39%: Calm/Content --- Pleased
Tellegen-Watson-Clark model: Q2-61%: Pleasant --- Happy/Joyful
PAD model: Range2-61%: Happy/Pleased-Pleasant
The percentage shown represents the highest number of participants that have
marked in the particular quadrant. So it is quite clear that all the three models
categorise the second set of parameters as happy, pleasant and joyful. For the PAD
model, most of the people marked pleasure, arousal and dominance as “medium” for
this set of parameters. Now considering the third set of parameters that is at A=1000
and V=300 for gesture1 of all three models, the results are:
©S.B.Sial, 2013
69
Gesture1: Point-point motion at V=2000 and A=200
Russell’s model :Q1-67%: Excited/Delighted ---- Aroused
Tellegen-Watson-Clarkmodel:Q1-50%:Alert/Delighted---Amazed/Surprized
PAD model: Range2-67%: Happy/Pleased --- Pleasant
The rest of the percentage is divided among all of the remaining categories, showing
a very small percentage falling in each of them. However, for this gesture only the
result shown by the PAD model differs from rest of the models. The other two
models place the emotion under the category of being excited, delightful, and alert.
According to the PAD model the perceived emotion comes under being happy or
pleased. However the people marked pleasure, arousal and dominance as “high” for
this set of parameters.
Similarly, shown below are the results for gesture2 (i.e. waving of the robotic arm)
for all the three set of parameters for all models.
Gesture2: Waving of robot at V=100 and A=15:
For Gesture2: Waving of robot at V=100 and A=5:
For Gesture2: Waving of robot at V=100 and A=1.5:
Russell’s model Tellegen-Watson-Clark model PAD model
Q1 0%
Q2 11%
Q3 83%
Q4 5%
Q1 5%
Q2 0%
Q3 12%
Q4 83%
R1 95%
R2 5%
R3 0%
Russell’s model Tellegen-Watson-Clark model PAD model
Q1 6%
Q2 33%
Q3 22%
Q4 39%
Q1 28%
Q2 61%
Q3 11%
Q4 0%
R1 17%
R2 78%
R3 5%
Russell’s model Tellegen-Watson-Clark model PAD model
Q1 72%
Q2 28%
Q3 0%
Q4 0%
Q1 67%
Q2 33%
Q3 0%
Q4 0%
R1 0%
R2 22%
R3 78%
©S.B.Sial, 2013
70
Gesture2: Waving of robot at V=100 and A=15
Russell’s model : Q3-83%: Tired/Sleepy --- Sad/Miserable
Tellegen-Watson-Clark model-83%: Q4: Unpleasant --- Sad/Tired
PAD model: Range1-95%: Sad/Tired --- Unpleasant
Gesture2: Waving of robot at V=100 and A=5
Russell’s model : Q4-39%: Calm/Content --- Pleased
Tellegen-Watson-Clark model: Q2-61%: Pleasant --- Happy/Joyful
PAD model: Range2-78%: Happy/Pleased --- Pleasant
Gesture2: Waving of robot at V=100 and A=1.5
Russell’s model : Q1-72%: Excited/Delighted ---- Aroused
Tellegen-Watson-Clarkmodel:Q1-67%:Alert/Delighted---Amazed/Surprized
PAD model: Range3-78%: Excited --- Aroused
This shows that for gesture2 the results for all the three set of parameters falls under
the same category from all the models. Moreover these results are similar and
support the ones obtained for Gesture1. Most of the people marked individual
parameters of PAD scale as “low, low, low”, “medium, medium, medium” and
“high, high, high” for the three sets of parameters respectively representing the three
emotions of being sad, happy and excited .
The results obtained for gesture 3 that is bowing down of the robotic arm for all
three set of parameters is given below:
Gesture3: Bowing down of robot at V=30 and A=30:
Russell’s model Tellegen-Watson-Clark model PAD model
Q1 0%
Q2 5%
Q3 83%
Q4 11%
Q1 5%
Q2 11%
Q3 12%
Q4 72%
R1 100%
R2 0%
R3 0%
©S.B.Sial, 2013
71
Gesture3: Bowing down of robot at V=30 and A=30:
For Gesture3: Bowing down of robot at V=50 and A=50:
For Gesture3: Bowing down of robot at V=100 and A=100:
Gesture3: Bowing down of robot at V=30 and A=30
Russell’s model : Q3-83%: Tired/Sleepy --- Sad/Miserable
Tellegen-Watson-Clark model: Q4-72%: Unpleasant --- Sad/Tired
PAD model: Range1-100%: Sad/Tired --- Unpleasant
Gesture3: Bowing down of robot at V=50 and A=50
Russell’s model : Q4-50%: Calm/Content --- Pleased
Tellegen-Watson-Clark model: Q2-78%: Pleasant --- Happy/Joyful
PAD model: Range2-83%: Happy/Pleased --- Pleasant
Gesture3: Bowing down of robot at V=100 and A=100
Russell’s model : Q1-61%: Excited/Delighted ---- Aroused
Tellegen-Watson-Clark-model:Q1-61%:Alert/Delighted--Amazed/Surprized
PAD model: Range3-61%: Excited --- Aroused
Russell’s model Tellegen-Watson-Clark model PAD model
Q1 0%
Q2 5%
Q3 83%
Q4 11%
Q1 5%
Q2 11%
Q3 12%
Q4 72%
R1 100%
R2 0%
R3 0%
Russell’s model Tellegen-Watson-Clark model PAD model
Q1 11%
Q2 17%
Q3 22%
Q4 50%
Q1 12%
Q2 78%
Q3 5%
Q4 5%
R1 17%
R2 83%
R3 0%
Russell’s model Tellegen-Watson-Clark model PAD model
Q1 61%
Q2 11%
Q3 0%
Q4 28%
Q1 61%
Q2 39%
Q3 0%
Q4 0%
R1 5%
R2 33%
R3 61%
©S.B.Sial, 2013
72
The first set parameter of velocity and acceleration is perceived as being sad,
unpleased etc. whereas the second and third are perceived as being happy and
excited respectively. The individual marking for pleasure, arousal and dominance on
the PAD scale is (low, low, low) for the first set of parameters which represents
being sad, for happy and pleasure mood the individual parameters are marked as
(high, low-medium, low) and for being alert and excited the individual P, A and D is
marked as (high, high, high).
6.6 Summary of the chapter
This chapter focuses on the techniques and methods that were followed in order to
perform the experiments with 18 different participants. It explains how the
participants have marked the questionnaire for perceived emotions on three different
scales. Later the results for these experiments are collected in the form of bar graphs.
The statistical results of each gesture for all three scales are also discussed in this
chapter. The concluded results shows that majority of the participants at low
velocity and acceleration, for all the three gestures, have marked the perceived
emotions in the category of being sad, unpleasant or tired. Whereas at medium level
of velocity and acceleration the perceived emotion for all three gestures, according
to all three scales, was pleasant, happy or pleased. However for high values of
velocity and acceleration the perceived emotional behaviour falls under the category
of alert, delighted, amazed or excited.
The next chapter will discuss the conclusions that are drawn from these results. It
will highlight the major findings and limitation of this research. It will also focus on
the possible future work that can be done.
©S.B.Sial, 2013
73
CHAPTER 7
CONCLUSIONS AND RECOMMENDATIONS
7.1 Conclusion
From the results produced it can be concluded the slow motion is perceived as sad,
unhappy and unpleasant by the participants, the medium level motion parameters for
velocity and acceleration is perceived as happy, joyful and calm by the participants
and the emotion that participants have associated with the fast motion is of being
excited, alert and aroused. These motion parameters are therefore considered
important for the change in user’s perception of emotions (Saerbeck and Bartneck,
2010). Therefore with the change of speed and acceleration the emotional mood
changes from being sad to happy to excited. This develops a link between the
change in user perception of emotions by varying the motion parameters of velocity
and acceleration (Ian et al., 2005).
This kind of robotic embodiment that is considered as machine robot is capable of
conveying emotions without any android features such as face etc. (Beck et al.,
2013). Moreover it is observed that the noise produced by the robot changes with
change of emotional behaviour (Eun et al., 2009). When the robot was perceived to
be sad or unhappy the noise associated with it was very low. However as the
perceived emotion changed from sad to happy and then to excited as the noise
associated with robotic embodiment increased exponentially.
This research gives rise to several questions that remain to be answered e.g. in the
field of care and medication, are slow movements of a robot perceived as a sad
gesture or a careful gesture by the patient? For industrial purposes can these
emotional robots have the same efficiency and productivity rate as the ones used
now? Further research and investigation is therefore needed in this area in order to
incorporate these emotional robots in various important fields of life effectively.
©S.B.Sial, 2013
74
7.2 Recommendations regarding hardware
The following improvements can be made in this project regarding the hardware:
Use FPGA with a higher space capacity and more DMA channels for the
transfer of data between RT and FPGA VIs
Use a NI 9403 C Series 32-Ch, 5 V/TTL Bidirectional Digital I/O Module to
read all joint encoders
A gripper provided by FESTO and many other companies can be attached at
the end of Robolink to introduce more functionally and for extending the
core concept of project
7.3 Recommendations regarding LabVIEW programming
The code on the FPGA as well as RT side can be improved further by:
Code on FPGA side should be reduced by either shifting it to RT VI or by
reducing it so to have space for new concepts of programming
NI 9403 module could be introduced in the code for keeping the track of the
joint positions
7.4 Research limitations
It is important to highlight that the poses and gestures were deliberately selected to
be expressive for the user. However it was important from the aspect of developing
a movement that should be expressive and communicative to the user ( Beck et al.,
2013). This might had an effect on the results found in this research.
Moreover the participants should be blinded from the data for changing values of
motion parameters on the questionnaire. This can be considered biased in finding a
relationship between velocity and acceleration, and envisaged emotion.
The sequence of motion parameters as well as gestures should also be randomized as
this might be helpful in predicting the next emotion in line.
Another potential bias associated with this robotic embodiment is the noise that it
makes during its motion. This noise rises with the increase in values of motion
parameters. At low values of motion parameters the noise associated is less. As the
values of velocity and acceleration increases the noise gets louder. Therefore this
might help the user to identify the perceived emotions.
©S.B.Sial, 2013
75
7.5 Future work
This study did not consider the effect of changing the embodiment in same robot
category that is of machine robot, to see if the change of embodiment affects the
results or not. Therefore same experiments shall be performed on different
embodiment to see the effect.
Introduction of new gestures and emotions in the embodiment can also increase the
scope of research. Although the results are quite reasonable, the sample size of
participants is quite small. It should be increased for the generalizability of results.
Additionally the robot should be equipped with some kind of soundproof material
for the reduction of noise.
This research shall be performed on android robot to see if the perception of user
differs by changing the robot to android one. This will also check that whether the
research supports Uncanny Valley theory or not.
©S.B.Sial, 2013
76
References
1. Anandan, T., 2013. The End of Separation: Man and Robot as Collaborative
Coworkers on the Factory Floor. Robotic online. URL:
http://www.robotics.org/content-detail.cfm/Industrial-Robotics-Featured-
Articles/The-End-of-Separation-Man-and-Robot-as-Collaborative-Coworkers-
on-the-Factory-Floor/content_id/4140.
2. Bartneck, C., Reichenbach, J., Carpenter, J., 2006. Use of Praise and Punishment
in Human-Robot Collaborative Teams, in: The 15th IEEE International
Symposium on Robot and Human Interactive Communication, 2006. ROMAN
2006. Presented at the 15th IEEE International Symposium on Robot and Human
Interactive Communication, 2006. ROMAN 2006, pp. 177–182.
3. Beck, A., Hiolle, A. & Cañamero, L., 2013. Using Perlin Noise to Generate
Emotional Expressions in a Robot. CogSci 2013 THE ANNUAL MEETING OF
THE COGNITIVE SCIENCE SOCIETY, pp.1845–1850. URL:
http://mindmodeling.org/cogsci2013/papers/0343/paper0343.pdf.
4. Bizzi , E. et al., 1984. Posture control and trajectory formation during arm
movement. , 4(11), pp.2738–2744. URL:
http://web.mit.edu/bcs/bizzilab/publications/bizzi1994.pdf.
5. Blythe, P.W., Todd, P.M., Miller, G.F., 1999. How motion reveals intention:
Categorizing social interactions, in: Simple Heuristics That Make Us Smart,
Evolution and Cognition. Oxford University Press, New York, NY, US, pp.
257–285.
6. Boris , K., 2007. Pointer oriented object detection method, involves defining class
of objects as action and realizing and processing pointer on vocabulary of
computer system in natural speech during initialization of objects. URL:
http://www.google.com/patents/DE102006052141A1?cl=en.
7. Braitenberg, V., 1984. Vehicles: Experiments in Synthetic Psychology First.,
USA: MIT press paperback edition 1986. URL:
http://books.google.co.uk/books?id=7KkUAT_q_sQC&printsec=frontcover&sou
rce=gbs_ge_summary_r&cad=0#v=onepage&q&f=false.
©S.B.Sial, 2013
77
8. Breazeal, C., 2003. Toward sociable robots. Robotics and Autonomous Systems
42, 167–175.
9. Coan, J.A., Allen, J.J., 2007. Handbook of Emotion Elicitation and Assessment.
Oxford University Press.
10. Coolcore, 2009. CoolCore the Low Power IP Core Company. URL
http://www.coolcore.co.uk/Home.html (accessed 7.29.13).
11. Dautenhahn, K. et al., 2009. KASPAR - a minimally expressive humanoid robot
for human-robot interaction research. , (3-4), pp.369–397. URL:
http://homepages.feis.herts.ac.uk/~comqkd/Dautenhahn-etal-Revised-
CompleteDocument-web.pdf.
12. Dautenhahn, K., 2007. Socially intelligent robots: dimensions of human--robot
interaction. Philosophical Transactions of the Royal Society B: Biological
Sciences, 362, pp.679–704. Available at:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2346526/pdf/rstb20062004.pdf.
13. Dautenhahn, K., 1999. ROBOTS AS SOCIAL ACTORS:AURORA AND THE
CASE OF AUTISM. In Third Cognitive Technology Conference CT’99, August,
San Francisco. URL:
http://www.gel.usherbrooke.ca/crj/Documentation/ctkerstin.pdf.
14. Dick, P.K., 2013. android project. Philip K. Dick Android project. URL
http://www.pkdandroid.org/ (accessed 7.26.13).
15. Ertugrul, N., 2000. Towards Virtual Laboratories a Survey of LabVIEW-
based teaching/learning tools and future trends. International Journal of
Engineering Education, 2000 pp. 1–10.
16. Eun, S.J. et al., 2009. Sound Production for the Emotional Expression of Socially
Interactive Robots. In Advances in Human-Robot Interaction. URL:
http://cdn.intechopen.com/pdfs/6453/InTech
Sound_production_for_the_emotional_expression_of_socially_interactive_robot
s.pdf.
17. Flanagan, J.R. & Ostry, D.J., 1990. Trajectories of human multi-joint arm
movements: Evidence of joint level planning . In The First International
Symposium on Experimental Robotics I. URL:
http://brain.phgy.queensu.ca/flanagan/papers/FlaOst_ERO_90.pdf.
©S.B.Sial, 2013
78
18. Folea, S., 2011. LabVIEW-practical applications and solutions. InTech, URL
http://www.intechopen.com/books/practical-applications-and-solutions-using-
labview-software (accessed 7.26.13).
19. Fong, T., Nourbakhsh, I. & Dautenhahn, K., 2003. A survey of socially
interactive robots. , Volume 42(Issues 3–4), p.Pages 143–166. Available at:
http://infoscience.epfl.ch/record/30017/files/CMU-RI-TR-02-29.pdf.
20. Fontys, 2013. IGUS robotic arm. Mechatronica. URL:
http://lectoraatmechatronica.wikispaces.com/Igus-robotic-arm (accessed 7.26.13).
21. Forlizzi, J. & DiSalvo, C., 2006. Service Robots in the Domestic Environment:
A Study of the Roomba Vacuum in the Home. In HRI ’06 Proceedings of the
1st ACM SIGCHI/SIGART. Human Robot Interaction. pp. 258–265. URL:
http://133.11.9.3/~takeo/course/2007/media/papers/roomba_at_home.pdf.
22. Gaertner, S. et al., 2010. Generation of Human-like Motion for Humanoid
Robots Based on Marker-based Motion Capture Data. In Robotics (ISR), 2010
41st International Symposium on and 2010 6th German Conference on Robotics
(ROBOTIK). Munich, Germany, pp. 1 – 8. URL:
http://ieeexplore.ieee.org/xpl/articleDetails.jsp?reload=true&arnumber=5756898
&tag=1.
23. Gaveau, J., Papaxanthis, C., 2011. The Temporal Structure of Vertical Arm
Movements. Open I beta. URL
http://openi.nlm.nih.gov/detailedresult.php?img=3134452_pone.0022045.g002&r
eq=4 (accessed 8.2.13).
24. Gizmag, team, 2002. Tiny lamprey-inspired robot could locate diseases
inside human body . URL http://www.gizmag.com/lamprey-robot-disease-
detection-epsrc/22016/ (accessed 7.26.13).Goodrich, M.A., Schultz, A.C., 2007.
25. Goodrich, M.A., Schultz, A.C., 2007. Human-robot interaction-A survey.
Foundations and Trends R © in Human–Computer Interaction Vol. 1, No. 3
(2007) 203–275.
26. Haddadin, S. et al., 2011. Towards the Robotic Co-Worker. In 14TH
International Symposium of Robotics Research. ISRR. Lucerne, Switzerland, pp.
261–282. URL: http://www.phriends.eu/isrr_09b.pdf.
©S.B.Sial, 2013
79
27. Heider, F., Simmel, M., 1944. An Experimental Study of Apparent Behavior.
The American Journal of Psychology 57, 243.
28. Höök, K., 2013. Affective Computing. The Encyclopedia of Human-Computer
Interaction, 2nd Ed. URL /encyclopedia/affective_computing.html (accessed
7.29.13).
29. Ian , B., Debayan , G. & Fei , H., 2005. Effects of low level changes in motion
on human perception of robots. In Development and Learning, 2005.
Proceedings. . The 4th International Conference on Development and Learning.
pp. 1–5. URL: http://work.debayangupta.com/smooth/BGH.pdf.
30. IGUS, 2013. Robolink. URL http://www.igus.eu/_wpck/pdf/global/DE-
EN_robolink_04-2013_s.pdf (accessed 7.26.13).
31. IGUS, 2013. IGUS plastics for longer life. IGUS. URL
http://www.igus.co.uk/wpck/default.aspx?PageNr=7904&C=US&L=en (accessed
7.26.13).
32. IGUS, 2012a. Robolink joint kit-documentation. URL
http://www.igus.eu/_wpck/pdf/global/Documentation_robolink_en.pdf (accessed
7.26.13).IGUS, 2012b.
33. IGUS, 2012b. Stepper motor. URL
http://www.igus.eu/_wpck/pdf/global/Motordatasheet_EN.pdf (accessed 7.26.13).
34. Jason , M.O., 2007. A theory for comparing robot systems. University of Illinois.
URL: http://www.cse.sc.edu/~jokane/pubs/OKa07.html.
35. Jie Cao, H.W., 2008. PAD Model Based Facial Expression Analysis.
Springer-Verlag Berlin Heidelberg pp.450–459. URL
http://hcsi.cs.tsinghua.edu.cn/accenter/paper/ac2005/assist/Papers2008/09-
PAD%20Model%20Based%20Facial%20Expression%20Analysis.pdf (accessed
7.29.13).
36. Kensinger, E.A., 2004. Remembering emotional experiences: The contribution
of valence and arousal. Reviews in the Neurosciences pp. 241–253. URL
https://www2.bc.edu/~kensinel/Kensinger_RevNeurosci04.pdf
37. Koeppe, R. et al., 2003. Robot-Robot and Human-Robot Cooperation in
Commercial Robotics Applications. In Robotics Research, The Eleventh
©S.B.Sial, 2013
80
International Symposium. ISRR. Siena, Italy.URL:
http://link.springer.com/chapter/10.1007%2F11008941_22#page-1.
38. Kuhlman, R., 2013. Introduction to the LabVIEW FPGA Module.URL
https://ni.adobeconnect.com/_a56821929/p28532990/?launcher=false&fcsConten
t=true&pbMode=normal (accessed 7.29.13).
39. Li, X., Zhou, H., Song, S., Ran, T., Fu, X., 2005. The Reliability and Validity
of the Chinese Version of Abbreviated PAD Emotion Scales, in: Tao, J., Tan, T.,
Picard, R.W. (Eds.), Affective Computing and Intelligent Interaction, Lecture
Notes in Computer Science. Springer Berlin Heidelberg, pp. 513–518.
40. Matsumaru, K., 2009. Discrimination of emotion from movement and addition
of emotion in movement to improve human-coexistence robot’s personal
affinity. In Robot and Human Interactive Communication, 2009. RO-MAN
2009. The 18th IEEE International Symposium on. Toyama, pp. 387 – 394.URL:
http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=5326345&tag=1.
41. Mehrabian, A., 1977. Nonverbal Communication. Transaction Publishers.
URL:http://books.google.co.uk/books?hl=en&lr=&id=XtYALu9CGwC&oi=fnd
&pg=PR7&dq=nonverbal+communication+by+albert+mehrabian&ots=5xNeQi8
kow&sig=G3O1zsT8yRfuio3oweb1XbkYvSw#v=onepage&q=nonverbal%20co
mmunication%20by%20albert%20mehrabian&f=false (accessed 7.29.13).
42. Mehrabian, A., 1980. Basic Dimensions for a General Psychological Theory:
Implications for Personality, Social, Environmental, and Developmental Studies.
Oelgeschlager, Gunn & Hain.
43. Mehrabian, A., 1996. Pleasure-arousal-dominance: A general framework for
describing and measuring individual differences in Temperament. Current
Psychology 14, 261–292.
44. Menzel, P., 2013. Kismet. Kismet. URL
http://www.ai.mit.edu/projects/humanoid-robotics-group/kismet/kismet.html
(accessed 7.26.13).
45. Mori, M., 2005. The Uncanny Valley. URL
http://www.androidscience.com/theuncannyvalley/proceedings2005/uncannyvalle
y.html (accessed 7.26.13).NASA, 2013. Robonaut.
46. NASA. URL robonaut.jsc.nasa.gov/default.asp (accessed 7.26.13).
©S.B.Sial, 2013
81
47. NI, 2013. About virtual instrumentation. National Instruments. URL
http://www.ni.com/white-paper/2964/en/ (accessed 7.29.13).
48. NI, 2010. Operating instructions and specifications CompactRIO cRIO-
9072/3/4 National Instruments. URL
http://www.ni.com/pdf/manuals/374639e.pdf (accessed 7.29.13).
49. NI, 2012. Integrated 400MHz Real-Time Controller and 2M gate FPGA.
National Instruments. URL http://sine.ni.com/nips/cds/print/p/lang/en/nid/203964
(accessed 7.29.13).
50. NI, 2012e. Advanced Motion Using NI CompactRIO. National Instruments.
URL http://zone.ni.com/wv/app/doc/p/id/wv-873 (accessed 7.29.13).
51. NI, 2013f. Building an NI Motion Control System. National Instruments.
URL http://www.ni.com/white-paper/12127/en/ (accessed 7.29.13).
52. NI, 2011b. CompactRIO Integrated Systems with Real-Time Controller and
Reconfigurable Chassis. National Instruments. URL
http://sine.ni.com/ds/app/doc/p/id/ds-204/lang/en (accessed 7.29.13).
53. NI, 2013e. FPGA Design, Development and Programming Tutorial. National
Instruments. URL http://www.ni.com/white-paper/3358/en/ (accessed 7.29.13).
54. NI, 2013d. FPGA Fundamentals. National Instruments. URL
http://www.ni.com/white-paper/6983/en/ (accessed 7.29.13).
55. NI, 2013. Hands-On: Introduction to LabVIEW Real-Time. URL
ftp://ftp.ni.com/pub/branches/uk/nidays2008/track_6/intro_to_lv_realtime.pdf
(accessed 7.29.13).
56. NI, 2013b. Increase your organization productivity with LabVIEW. National
Instruments. URL http://www.ni.com/white-paper/3494/en/ (accessed 7.29.13).
57. NI, 2011a. LabVIEW for measurement and data analysis. National
Instruments. URL http://www.ni.com/white-paper/3566/en/ (accessed 7.26.13).
58. NI, 2012b. Measurement and automation explorer help .National
Instruments. URL
http://digital.ni.com/manuals.nsf/websearch/A20E4510ACC3BADE862579FF00
789571 (accessed 7.29.13).
©S.B.Sial, 2013
82
59. NI, 2012d. NI 9401, 8 Ch, 5 V/TTL High-Speed Bidirectional Digital I/O
Module. National Instruments. URL
http://sine.ni.com/nips/cds/print/p/lang/en/nid/208809 (accessed 7.29.13).
60. NI, 2012c. NI 9501. National Instruments. URL
http://sine.ni.com/ds/app/doc/p/id/ds-292/lang/en (accessed 7.29.13).
61. NI, 2012a. NI cRIO 9074. National Instruments. URL
http://sine.ni.com/nips/cds/print/p/lang/en/nid/203964 (accessed 7.26.13).
62. NI, 2013c. NI LabVIEW Real-Time Module. National Instruments. URL
http://www.ni.com/labview/realtime/ (accessed 7.29.13).
63. NI, 2012f. Overview of the LabVIEW Robotics Module. National
Instruments. URL
http://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=
0CDYQFjAB&url=http%3A%2F%2Fwww.ni.com%2Fwhite-
paper%2F11564%2Fen%2Fpdf&ei=gkXtUei6PILTPOePgLgD&usg=AFQjCNG
s8l4u4paGnD2cZWTpUev7a06PjQ&bvm=bv.49478099,d.d2k (accessed
7.29.13).
64. NI, 2013a. What is CompactRIO. National Instruments. URL
http://www.ni.com/compactrio/whatis/ (accessed 7.29.13).
65. NI, 2013b. Designing Serial Arms (Robotics Module). National Instruments.
URL:http://zone.ni.com/reference/en-XX/help/372983D-
01/lvrobogsm/robo_arm_definition/.
66. Nishida, T., Jain, L.C., Faucher, C., 2010. Modelling Machine Emotions for
Realizing Intelligence - Foundations and Applications. URL
http://www.springer.com/engineering/computational+intelligence+and+complexit
y/book/978-3-642-12603-1 (accessed 9.06.13).
67. Posada-Gomez, R., Osvaldo, O., Martinez, A., Portillo-Rodriguez, O., Alor-
Hernandez, G., 2011. Digital Image Processing Using LabView, in: Folea, S.
(Ed.), Practical Applications and Solutions Using LabVIEW Software. InTech.
68. Posner, J., Russell, J.A., Peterson, B.S., 2005. The circumplex model of
affect: An integrative approach to affective neuroscience, cognitive development,
and psychopathology. Dev Psychopathol 17, 715–734.
©S.B.Sial, 2013
83
69. Qassem, M.A., Abuhadrous, I. & Elaydi, H., 2010. Modeling and Simulation of
5 DOF educational robot arm. In Advanced Computer Control (ICACC), 2010
2nd International Conference. Shenyang, pp. 569 – 574. URL:
http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=5487136&url=http%3A
%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D5487136
70. Remington, N.A., Fabrigar, L.R., Visser, P.S., 2000. Re-examining the
circumplex model of affect. Journal of Personality and Social Psychology 79,
286–300.
71. Russell, J.A., 1980. A circumplex model of affect. Journal of Personality and
Social Psychology 39, 1161–1178.
72. Russell, J.A., Lewicka, M., Niit, T., 1989. A cross-cultural study of a
circumplex model of affect. Journal of Personality and Social Psychology 57,
848–856.
73. Saerbeck, M., Van Breemen, A.J.N., 2007. Design guidelines and tools for
creating believable motion for personal robots, in: The 16th IEEE International
Symposium on Robot and Human Interactive Communication, 2007. RO-MAN
2007. Presented at the 16th IEEE International Symposium on Robot and Human
interactive Communication, 2007. RO-MAN 2007, pp. 386–391.
74. Saerbeck, M., Bartneck, C., 2010. Perception of affect elicited by robot
motion, in: Proceedings of the 5th ACM/IEEE International Conference on
Human-robot Interaction, HRI ’10. IEEE Press, Piscataway, NJ, USA, pp. 53–60.
75. Sandin, P.E., 2003. Robot Mechanisms and Mechanical Devices Illustrated.
McGraw-hill companies.
76. Tellegen, A., Watson, D., Clark, L.A., 1999. On the Dimensional and
Hierarchical Structure of Affect. Psychological Science 10, 297–303.
77. Thalmann, D., Ahn, J., Silvestre, Q., 2010. Asymmetrical Facial Expressions
based on an Advanced Interpretation of Two-dimensional Russells Emotional
Model. proceedings of ENGAGE 2010.
78. Thrun, S., Bennewitz, M., Burgard, W., Cremers, A.B., Dellaert, F., Fox, D.,
Hahnel, D., Rosenberg, C., Roy, N., Schulte, J., Schulz, D., 1999. MINERVA: a
second-generation museum tour-guide robot, in: 1999 IEEE International
Conference on Robotics and Automation, 1999. Proceedings. Presented at the
©S.B.Sial, 2013
84
1999 IEEE International Conference on Robotics and Automation, 1999.
Proceedings, pp. 1999–2005 vol.3.
79. Travis, J., Kring, J., 2013. LabView for everyone, 3rd ed. Prentice Hall PTR.
80. Trohidis, K., Tsoumakas, G., Kalliris, G., Vlahavas, I., 2011. Multi-label
classification of music by emotion. EURASIP Journal on Audio, Speech, and
Music Processing 2011, 4.
81. Tuomas, Eerola & Jonna, K. Vuoskoski, 2011. A comparison of the discrete and
dimensional models of emotion in music. , 39, pp.18–49. URL:
http://pom.sagepub.com/content/39/1/18.abstract.
82. Uno, A., 2009. RIBA robot nurse bear. Pink tentacle. URL
http://pinktentacle.com/2009/08/riba-robot-nurse-bear/ (accessed 7.26.13).
83. Watson, D., Clark, L.A., Tellegen, A., 1988. Development and validation of
brief measures of positive and negative affect: The PANAS scales. Journal of
Personality and Social Psychology 54, 1063–1070.
84. Wondolowski, C., Davis, D.K., 1991. The Lived Experience of Health in the
Oldest Old: A Phenomenological Study. Nurs Sci Q 4, 113–118.
85. Yang, dan, Lee, W., 2004. Disambiguating music emotion using software
agents. Presented at the 5th International Conference on Music Information
Retrieval, Universitat Pompeu Fabra, Barcelona, Spain, pp. 10–14.
©S.B.Sial, 2013
85
Appendices
Appendix A: Drawings for IGUS robotic arm
Appendix B: Tube length of joints
Appendix C: Technical data for tube lengths
Appendix D: Datasheet of integrated Hall IC’s and
configuration of sensor lines
Appendix E: Technical data for stepper motors
Appendix F: Complete specifications for drive unit
Appendix G: Technical Datasheet of NI 9501
Appendix H: Technical Datasheet of cRIO 9074
Appendix I: Technical Datasheet of NI 9401
Appendix J: Properties of CompactRIO 9074
Appendix K: Connection of motors with 9501
Appendix L: Wiring of chassis and cRIO
Appendix M: Mechanical parts and joint types of Robolink
Appendix N: Ethical approval and consent form
Appendix O: LabVIEW Code (CDROM)
©S.B.Sial, 2013
86
Appendix A: Drawings for IGUS robotic arm
©S.B.Sial, 2013
87
©S.B.Sial, 2013
88
©S.B.Sial, 2013
89
Appendix B: Tube length of joints
©S.B.Sial, 2013
90
Appendix C: Technical data for tube lengths
©S.B.Sial, 2013
91
©S.B.Sial, 2013
92
Appendix D: Datasheet of integrated Hall IC’s and
configuration of sensor lines
©S.B.Sial, 2013
93
©S.B.Sial, 2013
94
©S.B.Sial, 2013
95
Configuration sensor lines for pivoting movement
Configuration sensor lines for rotating movement
+5V Red
GND Black
Hall-Sensor White
Encoder Index Green
Encoder Channel A Blue
Encoder Channel B Yellow
+5V Red/Blue
GND Brown
Hall-Sensor Grey
Encoder Index Grey/Rose
Encoder Channel A Violet
Encoder Channel B Rose
©S.B.Sial, 2013
96
Appendix E: Technical data for stepper motors
©S.B.Sial, 2013
97
©S.B.Sial, 2013
98
©S.B.Sial, 2013
99
©S.B.Sial, 2013
100
©S.B.Sial, 2013
101
©S.B.Sial, 2013
102
©S.B.Sial, 2013
103
Appendix F: Complete specifications for drive unit
©S.B.Sial, 2013
104
©S.B.Sial, 2013
105
©S.B.Sial, 2013
106
Colour coding of rope
Motor order & joints Joints Rope colours
Motor1: joint1 Rotational Red
Motor2: joint3 Rotational White
Motor3: joint3 Pivot Yellow
Motor4: joint2 Rotational Blue
Motor5: joint2 Pivot Black
Rope channels to implement movements
The internal diameter of the rope is 50mm. Rope channel for the pivoting and
rotation movement are as follow (IGUS, 2013.):
90°=π x d/4=~ 39mm
180°=π x d/2=~ 79mm
360°=π x d=~ 160mm
Dimension of multi-axis joint
The dimensions of multi-axis joint of robolink that is used in this project are as
shown below in the figure (Fontys, 2013):
©S.B.Sial, 2013
107
Appendix G: Technical Datasheet of NI 9501
©S.B.Sial, 2013
108
©S.B.Sial, 2013
109
©S.B.Sial, 2013
110
Appendix H: Technical Datasheet of cRIO 9074
©S.B.Sial, 2013
111
©S.B.Sial, 2013
112
©S.B.Sial, 2013
113
©S.B.Sial, 2013
114
©S.B.Sial, 2013
115
©S.B.Sial, 2013
116
©S.B.Sial, 2013
117
©S.B.Sial, 2013
118
Appendix I: Technical Datasheet of NI 9401
©S.B.Sial, 2013
119
©S.B.Sial, 2013
120
©S.B.Sial, 2013
121
©S.B.Sial, 2013
122
©S.B.Sial, 2013
123
©S.B.Sial, 2013
124
Appendix J: Properties of CompactRIO 9074
General properties of cRIO 9074
FPGA properties for cRIO 9074
Properties of chassis for cRIO 9074
Physical specifications for cRIO
Product Name cRIO-9074
Form Factor CompactRIO
Product Type Controller(Computing Device)
Part Number 779999-01
Operating System/Target Real-Time
LabVIEW RT Support Yes
CE Compliance Yes
FPGA Spartan-3
Gates 2000000
Number of Slots 8
Integrated Controller Yes
Input Voltage Range 19V, 30V
Recommended Power Supply: Power 48W
Recommended Power Supply: Voltage 24V
Power Consumption 20W
Length 28.97 cm
Width 8.81 cm
Height 5.89 cm
Weight 929 gram
Minimum Operating Temperature -20 °C
Maximum Operating Temperature 55 °C
Maximum Altitude 2000 m
©S.B.Sial, 2013
125
Appendix K: Connection of motors with 9501
In order to change the direction of motion for the joint, just reverse the cables for
that specific joint. The table below tells about the sequence of motors for the each
joint of this platform and also represents the type of joint that’s rotational or pivot.
Motor and joints representation
Sequence of connection for motors with 9501
Motor number Slot number for 9501 on chassis
1 1
2 5
3 4
4 3
5 2
Motor number Joint number Type of joint
1 1 Rotational
2 3 Rotational
3 3 Pivot
4 2 Rotational
5 2 Pivot
©S.B.Sial, 2013
126
Appendix L: Wiring of chassis and cRIO
Wiring power to chassis
The external power supply is required to be connected with the chassis of cRIO
9074. This will give power to all the chassis slots. It is also provided with the
reverse-voltage protection. Following steps should be followed for connecting
power supply with the chassis:
o The COMBICON connector shown should be tightened in the chassis with
the help of screw provided at its both ends.
o Connect the positive (red) lead of power supply with the upper most terminal
V1 and the negative (black) lead with the lower terminal C of the
COMBICON.
o Insert the wires and tighten them with the screws available at the top of
connector.
Power on cRIO 9074
When the chassis is first powered on, there will be two lights blinking, the power
and the status lights. It is important to understand the LED indication on the chassis
to avoid any problems during working.
LED indication on cRIO 9074
The LED indications are stated below:
COMBICON (NI, 2010)
LEDs indication for cRIO 9074 (NI, 2010)
©S.B.Sial, 2013
127
o The power LED is lit when the cRIO 9074 is powered on indicating that the
correct voltage is being supplied to the chassis
o The second LED is FPGA LED which is used for debugging of the
application. To define these LED FPGA mode or NI RIO mode is used
o The third LED is the status LED. As the name tells that this LED indicates
the status of the chassis. If this LED is off it means that it is under normal
operation. The blinking of this LED indicates an error condition. The type of
error then depends on the manner of blinking. The following table discusses
the manner in which this LED flashes by relating it with the error condition.
o The user1 LED can be defined by the user depending upon the application.
To define this, use RT LEDs in LabVIEW. However in our project this LED
is not used.
LEDs indication for error condition (NI, 2012b)
Flashes
Every Few
Seconds
Indication
1
The chassis is un-configured. Use MAX to configure the chassis. Refer to the
Measurement & Automation Explorer Help for information about configuring
the chassis
2
The chassis has detected an error in its software. This usually occurs when an
attempt to upgrade the software is interrupted. Reinstall software on the
chassis. Refer to the Measurement & Automation Explorer Help for
information about installing software on the chassis
3
The chassis is in safe mode because the SAFE MODE DIP switch is in the
ON position or there is no software installed on the chassis
Continuously
flashing
The chassis has detected an unrecoverable error. Contact National Instruments
continuously
flashing or
solid
The device may be configured for DHCP but unable to get an IP address
because of a problem with the DHCP server. Check the network connection
and try again. If the problem persists, contact National Instruments
©S.B.Sial, 2013
128
Reset option for chassis
The table below tells about the reset options for the chassis available on cRIO 9074.
This explains how it will behave on giving a reset. RIO devices setup utility is used
for the reset option.
Configuring settings and obtaining IP for target
Various steps below indicate how to connect the hardware with the LabVIEW
project (NI, 2012b):
1. Connecting with Ethernet/cross over cable:
The first step in connecting the hardware with the LabVIEW software is to connect
the chassis to a network. This can be done by connecting the chassis with the
Ethernet network using RJ-45 Ethernet port 1 on the front panel of the controller as
shown in Fig 4.6. A crossover cable can also be used for connecting directly to a
computer or laptop. The Ethernet cable is used for the communication of the host
computer and the chassis. If they are not connected through some IP address they
can communicate with each other through cross over cable. However for prior case
the subnet for the host computer and the chassis should be same.
2. LabVIEW Real-time booting:
Before getting onto the configuration settings, it is important to make sure that the
system is being booted in real time. There are several different boot modes that are
available and are listed below:
Normal boot mode
IP reset
Start-up application disabled
Safe mode
Chassis Reset
Option
Behaviour
Do not auto load
VI
Does not load the FPGA bit stream from flash memory
Auto load VI on
device power up
Loads the FPGA bit stream from flash memory to the FPGA
when the controller powers on
Auto load VI on
device reboot
Loads the FPGA bit stream from flash memory to the FPGA
when you reboot the controller either with or without cycling
power.
Reset option for cRIO 9074 (NI, 2012b)
©S.B.Sial, 2013
129
Uninstall mode
The real time target that is being used decides the booting of target. There are
various helps documents available online for booting of the hardware system in real-
time.
3. Configuring network by MAX:
The hardware/remote system that is being used must be assigned with a certain IP
address for it to communicate with the host PC. For configuring network settings,
the device is detected in Measurement and Automation Explorer (MAX). In MAX
there is a tab of Remote System at the left side. By expanding that tab, the selection
of particular remote system is done in order to configure it.
It can be seen in the figure above that by expanding the remote targets; the particular
device cRIO 9074 that is being used in the project is detected.
The IP address can be seen as 169.254.84.198. Once the IP is obtained the second
thing is installing the software on the remote device. For this, expand the particular
target on which the software has to installed, right click and select an option of
add/remove software. Select the software that is required to be installed on the
devices and start installing. After the installation is complete, on expanding software
tab one can see different software installed as in figure below:
Measurement and automation explorer
©S.B.Sial, 2013
130
.
Configuring LabVIEW project with hardware:
1. Once the software is installed on the target device, the chassis is mounted
with all the input and output modules and the controller is powered up, the next step
is adding the FPGA real-time device into the LabVIEW project to start with the
graphical programming interface.
For this, create a new project in LabVIEW, and right click my computer under
project explorer and select New>>Target and devices and shown in the figure
below:
Different software’s installed on remote device
©S.B.Sial, 2013
131
2. The installed FPGA devices will be shown when existing target/devices
option is clicked. Expand the Real-Time CompactRIO option to find the particular
controller under that as shown in the figure below:
Adding target device in LabVIEW project
Existing target and devices
©S.B.Sial, 2013
132
3. Select the specific controller and press ok. Another window will appear
asking for the mode of interface. Select LabVIEW FPGA interface that enables to
use C-series modules from LabVIEW FPGA VI’s.
4. On pressing continue after selecting the mode, the next step will be
discovering of the chassis. The software will ask whether to discover or not. Then
press discover C series modules as in figure:
5. This process of discovering the modules will take few seconds. After that
they will be automatically added in the main project as shown below:
Selecting programming mode
Discovering the modules
©S.B.Sial, 2013
133
6. At this stage the chassis, FPGA controller and all the drivers have been added
into the project successfully. Now it is time to start with the programming of RT and
FPGA VI’s. Right click the chassis and click create new VI. That will be the RT VI.
And for FPGA VI right click the FPGA target in the project and then build a new VI
under that.
Modules in project file
©S.B.Sial, 2013
134
Appendix M: Mechanical parts and joint types of Robolink
Mechanical parts of Robolink
There are several different components of robotic arm that are also shown in fig: 3.1
above. These include (IGUS, 2012a):
Joint types
Articulated arm
Angle sensors
Actuators
Draw wires
Stepper motors
Drive units
Accessories
Controls
The dimensional drawings and the exploded drawings of Igus robotic arm are
attached in the appendix A. These parts are discussed below:
Joint types
There are four types of joint types that are offered by Igus with this robotic arm
(IGUS, 2012a) namely: Swivel joint, rotating joint, Symmetric joint, Asymmetric
joint. The specifications for each of them are discussed below:
Type, version and joint limitation for IGUS robotic platform (IGUS, 2013)
Joint type Version of joint Range of joint
Swivel Joint
RL-50-PL1
+/- 90° swivel range
Rotating Joint
RL-50-TL1
+/- 90° rotating range
Symmetric Joint
(2 axes joint)
RL-50-001
+/- 90° swivel range
+/- 270° rotating range
Asymmetric Joint
(2 axes joint)
RL-50-002
+130/- 50° swivel range
+/- 270° rotating range
©S.B.Sial, 2013
135
These joints are provided with mechanical limits, but these can be removed.
Rotating and pivoting joints, are shown below in the figures
.
Figure a) represents the symmetric 2 axis joint and Figure b) represents the
Asymmetric 2 axis joint. The version of upper joint in Figure b) is RL-50-001, and
the pivot angle allowed is +/-90°. For lower joint the version is RL-50-002 and the
pivot angle allowed is +130°/-50°.
Articulated arms
With the help of these four joints, one can make one’s own customized arm by
choosing the link lengths and the type of material required for the link. There are
basically three types of material available for the links:
Aluminium tubes
Fiberglass tubes
Carbon-fibre tubes
Depending upon the utility the type is chosen. Standard tubes are made up of
aluminium with a diameter of 26mm and the tube length can be customized by the
user. The standard length of the link is 100mm and the tubes are hollow and with an
Pivot Rotating
Symmetric Asymmetric
(a) Pivoting and Rotating joints (IGUS, 2012a)
(b) Symmetric and Asymmetric joints (IGUS, 2012a)
©S.B.Sial, 2013
136
interior contour. The purpose of this contour is to avoid the rotation on joint
interface. Figure a) shows the interior contour of the connecting tubes used for the
Robolink. Figure b) represents different types of materials that can be used for
Robolink FGC, aluminium and CFC are shown respectively. Other specification like
length of the tubes, visible tube length and rotating point distance are attached in the
appendix B.
DOF of an articulated arm
The four types of joints results in 31 different types of configuration for the
articulated arm. The versions can be configured for 1-5 degrees of freedom. The
figures below represents articulated arms with different degrees of freedom.
Inside of tube and Types of material (IGUS, 2012a)
Articulated arm with 1 DOF
(IGUS, 2012a) Articulated arm with 2 DOF
(IGUS, 2012a)
Articulated arm with 3 DOF (IGUS, 2012a)
(a) (b)
©S.B.Sial, 2013
137
The technical data for these articulated arms from 01-31with a link length of 100mm
are attached in the appendix C. All articulated arms can be ordered with option of
the angle sensors.
Angle sensors
The optional sensors provided with the Robolink are magnetic incremental encoders.
Each axis has a magnetic ring and a Hall sensor associated with it. The specification
of the magnetic rings is different for swivel and rotating motion. For the swivel
motion it has 31 pole pairs and one additional South Pole, whereas for rotating
motion it has 29 pole pairs in total and an additional pole. Figure below shows the
magnetic ring and sensor unit for two axis joints:
Articulated arm with 4 DOF (IGUS, 2012a)
Articulated arm with 5 DOF (IGUS, 2012a)
Magnetic rings for swivelling and rotating respectively (IGUS, 2012a)
©S.B.Sial, 2013
138
The resolution of this encoder per axis is as follow:
Swivel motion 4960 counts per revolution, resolution 0.073 degrees
Rotating motion 4640 counts per revolution, resolution 0.078 degrees
Each joint has incremental encoders. The specification of particular arm for pivoting
and rotation are given below (Fontys, 2013):
Pivoting
o 31 pole pairs
o 40 pulses/pole pairs
o 160 positions/pole pairs
o 1240 pulses/revolutions
o 4690 positions/revolutions
Rotation
o 29 pole pairs
o 40 pulses/pole pairs
o 160 positions/pole pairs
o 1160 pulses/revolutions
o 4640 positions/revolutions
Accuracy
o For pivoting: 0.0726 degrees
o For rotation: 0.0776 degrees
Hall sensor for rotation
Encoders for rotation
Encoders for pivoting
Hall sensor for pivoting
Sensor unit (IGUS, 2012a)
©S.B.Sial, 2013
139
The main function of the incremental encoder is to convert the angular position of
the joint into digital information. It provides information about the motion of the
joint and further information about speed, distance and position can also be deduced.
It has three digital outputs: A, B and Index. The pulses A and B are quadrature
outputs, as they are 90° out of phase from each. Figure below shows A, B and index
pulses for the encoder.
The datasheet for the Hall sensors and the configuration of sensor line that
represents the colour coding of wires for pivoting as well as rotating joints is
attached in appendix D.
End effectors
The Robolink system is not provided with end effectors, however one can connect
various end effectors to the last link of the system if required. There are various
pneumatic and electrical grippers recommended for this robotic arm. FESTO® and
SCHUNK® provide pneumatic and electrical grippers for these arms. A standard
adaptor is provided by IGUS® for these grippers. Figure below shows some of the
popular gripper used.
A/B and index signals for the encoder (IGUS, 2012a)
©S.B.Sial, 2013
140
In this project there is no gripper or effector. As the main focus of this project was
on the behavioural movement of robot, no end effector was required.
Draw wires
The system of driving joint movement is by draw wires that are generally made from
Dyneema® (IGUS, 2013). There is a special coating that ensures long life and less
friction of these wires. The specifications of the wire used are as follow (IGUS,
2012a):
12-strand braiding
Diameter=2mm
Breaking strength=3.500N
Operating elongation=1% approx.
These wires are held in the drive wheel with the help of a nipple crimped on to the
end can be seen in figure below:
This
nipple is fitted
into the drive wheels shown in fig (a) below, so as to hold it properly. The tension in
the wires should be adequate to avoid play in the joint. Typically it is between 5-
10N at idle (IGUS, 2012a). If the tension in the wires is too high, the working life is
reduced because of wear and friction. The robotic arm is provided with the tools to
adjust the tension of these wires.
The robotic arm consists of multiple joints that are combined together in series. All
the joints are independent from each other because of the sequence through which
wires are fed into these joints. There is a limitation for the number of joints that can
be added in series, because only 4 wires can be fed through the lower joint. The pre-
assembled structure for the wires passing through lower joint is shown in the figures
below:
Standard pneumatic grippers used (IGUS, 2012a)
Assembled drive wheel with wire and wire with nipple (IGUS, 2012a)
©S.B.Sial, 2013
141
Figure (a, b, c, d) shows the feeding of wires through the lower joint of articulated
arm platform. Fig (a) shows that two wire pairs are being fed through the lower
joint. Fig (b) shows upper connecting element for guiding the wire. Fig (c) shows
Bowden cable segments for parallel feed-through and Fig (d) shows lower
connecting element of the robotic arm (IGUS, 2012a).
Stepper motors
Igus® uses stepper motors as the drive system for this articulated arm. However
alternate drive systems are also possible to control the motion of this platform. The
features of the stepper motor used are as follow (IGUS, 2013.):
Two phase hybrid stepper motors that are bipolar
Comes with plug/stranded wires
It has an option of encoder/brake
(a, b, c, d): Feeding of wires through joints (IGUS, 2012a).
©S.B.Sial, 2013
142
The stepper motors used for this project is: MOT-AN-S-060-020-056-L-LAAAA.
The stepper motor used for this project has version AN, maximum DC voltage of
60V, holding torque is 20Nm, and flange dimension is 56mm (NEMA 23) with
motor connection of stranded wires. The technical drawings and data specifications
are attached in the appendix E. Table below shows the key data for these motors
Motor NEMA 23
Maximum voltage 60 VDC
Nominal voltage 24-48 VDC
Nominal current 4.2A
Holding torque 2 Nm
Distance over hubs 56mm
Specifications for stepper motor drives (IGUS, 2013.)
Product code layout for stepper motors (IGUS, 2013)
©S.B.Sial, 2013
143
Fig. below shows the stepper motor used in the project with stranded conductor
version. It has four coloured wires. Table below shows the colour coding of wires.
Drive modules
The stepper motors used are equipped with a planetary gear and a tensionable drive
wheel system by the manufacturers. The reason for this is to increase the torque of
motors. Table below shows the configuration of gears used (IGUS, 2012b):
Drive units
There is an option of assembling the drive modules into drive units. This makes the
whole mechanism assembled. This process of fitting is done at factory. This
Pins Colour Signals
1 Brown A/
2 White A
3 Blue B
4 Black B/
Colour representation of motor wires
Motors Reduction gearing
NEMA 23 1:16
Gear ratio of stepper motors (IGUS, 2012b)
Stepper motors with four stranded wires (IGUS, 2012a)
Complete drive module (IGUS, 2013)
©S.B.Sial, 2013
144
complete module consists of articulated arm and tested drive units by Igus®. The
complete assembly is shown Fig. below:
There are some user selectable options that are available while ordering this
complete system (IGUS, 2012b).
Selection of DOF: 3-5 degrees of freedom can be selected depending upon
the requirement
Selection of angle sensors: joints of the system can be selected with or
without the angle sensors.
Adjustable link lengths: the length of the links can be selected from 100mm
to 1000mm
Selection of motors: two options for motor selections are available: NEMA17
or NEMA23
End effectors can also be added depending upon the requirement like various
grippers etc.
The units with the complete specifications are in appendix F.
Accessories/Spare parts
There are many other accessories that are provided by Igus® depending upon the
requirement e.g. rope end fitting, tensionable drive wheels, flange shaft blocks, rope
guides outside/inside, guide rollers, rope tensioners and many more.
Complete system (IGUS, 2012b)
©S.B.Sial, 2013
145
Torque and force
Torque and force are important parameters to be considered as two different motions
are made i.e. rotational and pivoting. At Igus® the data below is available (Fontys,
2013):
For pivoting movement 12 Nm at force 600N on the rope
For rotating movement 5Nm at force 300N on the rope
Torque for pivoting movement +/-27.88 Nm
For rotating movement +/-15 Nm
Rope management
The whole system is driven by cables as discussed above. Figure above shows the
guiding arrangement of the cables. Different colours of cables are used for each
motor so that it is easy to distinguish that which joint is connected to which motor.
The details of the rope channels are in appendix F
Pictorial representation of motor to drive wheel
Figure below shows the whole assembly from motors to drive wheel
Motor Gear Coupling Camp Drive wheel
Rope guiding system used in arms (Fontys, 2013)
Motor to drive wheel (IGUS, 2013)
©S.B.Sial, 2013
146
Appendix N: Ethical approval and consent form
Ethical approval
©S.B.Sial, 2013
147
Consent form
©S.B.Sial, 2013
148